2.1 κ-Köhler theory and the ZSR mixing rule

The single-parameter approach to describing a particle's propensity to grow hygroscopically proposed by Petters and Kreidenweis (2007) was employed here. This combines all variables that depend on a particle's chemical composition into a single parameter, κ. The relationship between the κ value and the HGF is expressed by Eq. (1):

where RH is the relative humidity expressed as a fraction, HGF is the hygroscopic growth factor, D_d is the dry particle diameter and

where σ_s∕a is the surface tension at the solution–air interface, M_w is the molar mass of water, R is the universal gas constant, T is the absolute temperature and ρ_w is the density of water. This equation is solved iteratively.

The value of κ increases with increasing hygroscopicity. Thus, κ will be 0 for a non-hygroscopic particle, above 1 for more hygroscopic particles and up to around 1.5 for the most hygroscopic particles such as NaCl. Typically, values for continental ambient aerosol fall between 0.1 and 0.4 (Pringle et al., 2010). Values of κ used for the species considered here are listed in Table 1. If ambient aerosol is assumed to be internally mixed, the κ values of the different components can be combined to provide a total κ value for the mixed particle, using Eq. (3):

where ε_i represents the volume fraction of an individual chemical component. This follows from the ZSR mixing rule described by Zdanovskii (1948) and Stokes and Robinson (1966). Once the value of κ has been determined for a given particle and if the RH is known, the relationship shown in Eq. (1) can be used to calculate the particle's expected HGF.

Table 1Density, kappa values and refractive indices for aerosol species considered here.

^a Weast (1985), in Morgan et al. (2010). ^b Penner et al. (1998), in Morgan et al. (2010). ^c Lowenthal et al. (2003). ^d Stelson (1990), in Taylor et al. (2015). ^e Bond and Bergstrom (2006), in Morgan et al. (2010). ^f Liu et al. (2014). ^g Suda et al. (2012). ^h Weingartner et al. (1997), in Jurányi et al. (2010). ⁱ Toon et al. (1976), in Morgan et al. (2010). ^j Bond and Bergstrom (2006).

Download Print Version | Download XLSX

2.2 Ion pairing

The AMS provides a time series of nitrate (${\mathrm{NO}}_{\mathrm{3}}^{-}$), sulfate (${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$) and ammonium (${\mathrm{NH}}_{\mathrm{4}}^{+}$) fragments. However, quantities of complete neutral salts are needed in order for the ZSR mixing rule described above to be used. Here, moles of neutral salts were established using the ion pairing scheme outlined by Gysel et al. (2007), as shown in Eq. (4). Calculations for H₂SO₄ and HNO₃ are not included as the aerosol was found to be in charge balance in all cases considered here; the ZSR calculations returned zero for both of these compounds.

2.3 Optical properties

Aerosol optical properties were calculated using a Mie code, based on a set of algorithms developed by Yang (2003), which uses a supplied refractive index, wavelength of light and aerosol diameter to calculate properties including the extinction coefficient for a multi-layered sphere. The refractive index used here was calculated by considering the weighted sum of the refractive indices of the individual chemical compounds in a particle, again assuming volume mixing. This technique does not account for some effects, including possible lensing effects from the addition of water to particles or the co-condensation of other volatile compounds with water. The individual refractive indices of the components considered are shown in Table 1. The effect of absorbed water on the refractive indices was taken into account, with the refractive index for water of 1.33−0i being used.

3.1 Observations

Figure 2a shows the average chemical composition of submicron aerosol in the region outlined by the box in Fig. 1. Data have been averaged to 500 m bins. Figure 2b shows the relative proportions of each chemical species at each altitude bin. As can be seen from Fig. 2b, the chemical distribution of aerosol was reasonably constant in this region in the lower five bins, with the overall concentration decreasing with altitude. Organic aerosol contributed the largest proportion to the aerosol loading, making up over 50 % of the total mass. ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ was around 20 % in all cases. Black carbon was just under 15 %. Contributions from ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ were each less than 10 %. This proportional composition changes in the 3000 m bin, with ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ becoming the most prominent and the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ contribution decreasing. It is likely this reflects the composition of aerosol immediately above the boundary layer top. However, since this altitude was only flown once during the morning of 15 July, it cannot be determined how representative this measurement is of the composition in general. If this were to be included in the profile it would only enhance the extinction at the top of the boundary layer, although the absolute concentrations are very small and make a limited contribution to the total ambient AOD.

Figure 2(a) Average chemical composition of aerosol in the Savè region, as measured by the AMS and SP2 (black carbon). Results are displayed at standard temperature and pressure (STP) and altitude is above sea level (a.s.l.) The centre point indicates the median concentration and the bars the lower and upper quartiles. (b) The percentage contribution of each chemical species in each altitude bin. The 3000 m is shaded as data only exist at this altitude from 1 h on 15 July.

Download

Figure 3 shows the normalised average number and surface area size distributions of submicron aerosol in the region outlined by the box in Fig. 1, as measured by the SMPS on board the ATR aircraft. Two distinct size modes can be seen, indicating that the region contains smaller, fresher aerosol with a mode at around 60 nm and a large accumulation mode containing more aged aerosol. The smaller Aitken mode was highly variable, as can be seen by the large spread in the data in Fig. 3a. The accumulation mode, however, remained present and relatively stable (with an interquartile range of ±30 % of the median) throughout the campaign. An optical particle counter (Grimm, model 1.109) was present on the Twin Otter aircraft during the campaign and did not measure a significant coarse mode during these flights. We therefore limit this analysis to accumulation mode aerosol only and use the extrapolated measured SMPS distribution for the remainder of this work.

Aircraft observations during the DACCIWA campaign period showed little variation in total aerosol concentration as a result of the diurnal cycle (see Fig. 2). When the data were sorted according to the time of day they were observed and considered in 2-hourly time bins, the largest deviation in the bin means from the overall campaign mean was around 11 %. Comparing aerosol profiles from different times of day did not reveal any up- or downwards trend in the aerosol concentrations as a function of the time of day, although there was some variability in mass loadings from one day to the next. Similarly, there was no evidence in the data for a trend in the proportional distribution of aerosol species across the diurnal cycle. For example, the average organic-to-sulfate ratio varied by less than 10 % across the diurnal cycle. The boundary layer aerosol during the DACCIWA campaign was found to have been significantly influenced by long-distance transport from biomass burning in central and southern Africa, which is likely to account for the homogeneity in aerosol properties observed here. This phenomenon will be explored in future work. All DACCIWA flights took place between the hours of 07:00 and 19:00 and were therefore unable to provide information about potential changes in aerosol properties occurring during the night.

As the objective of this work is to examine the effects of high RH, and given the lack of sufficient evidence in this dataset suggesting a trend in aerosol concentrations over the diurnal cycle, a median aerosol profile and chemical composition are used here to explore the effects of changes in RH on the AOD. However, it is important to note that the resultant calculations only provide an indication of the effect of RH on the AOD and are unable to capture the true range of AOD that would result from including variations in the aerosol loading, which would scale the resulting AOD linearly.

Figure 4 shows RH measurements from radiosondes released from the ground site in Savè. These plots show the mean RH and standard deviations at every 100 m interval up to 3 km, averaged across four times of day. These data show a clear diurnal cycle in the RH, with a humid boundary layer during the night and early morning becoming drier towards noon and into the evening. These features are characteristic of the Gulf of Guinea maritime inflow (Adler et al., 2017; Deetz et al., 2018b), a coastal air mass that propagates inland in the evening, bringing cool air to the southern West African region. The increase of RH in the evening is mainly caused by cooling due to this cold air advection (Adler et al., 2019; Babić et al., 2018). The RH in the pre-frontal region is generally around 75 %, while post-frontal RH rises to be almost consistently above 90 % (Deetz et al., 2018a) and leads to low-level cloud formation in the night (Adler et al., 2019; Babić et al., 2018).

The dataset of RH measurements used here is valuable, as models are often unable to replicate the vertical RH profile (Hannak et al., 2017). This is due to the complex interplay between shallow and deep convection, which changes the vertical distribution of water vapour and clouds. These factors influence the surface energy budget, which can in turn create deep convection, in a feedback loop. The uptake of water by aerosol particles adds another layer of complexity to this picture.

The time series panel gives insight into the day-to-day variability in RH. Four phases of the 2016 West African monsoon have been defined by Knippertz et al. (2017) and are delineated by the vertical dashed lines in Fig. 4e. The first phase (pre-onset) was characterised by rainfall near the coast, which moved inland during the second phase (post-onset). This can be seen in the drier start to the period shown in Fig. 4e. This was followed by a period of higher humidity in phase 2, during which humidities rarely dropped below 80 %. The final week of the post-onset phase was slightly drier than those preceding it, which was associated with a cyclonic system that passed Savè on 16 July, transporting dry air northwards (Kalthoff et al., 2018). During phase 3, the rainfall maximum shifted again back to the coast and during phase 4, the recovery of the monsoon, the maximum was once again found inland.

Figure 3The (a) number size distribution and (b) volume size distribution of submicron aerosol below 2500 m in the area outlined by the box in Fig. 1. The dark blue lines show the median size distributions; the darker shading represents data between the 25th and 75th percentiles (50 % of the data) and the lighter shading data between the 10th and 90th percentiles (80 % of the data). The thick black line represents a log-normal curve fitted to the accumulation mode. The rapid drop-off at the high end of the surface area distribution is an artefact of the SMPS size limit.

Download

Figure 4The mean and standard deviation of relative humidity measurements with altitude at (a) 06:00 UTC (n=46), (b) 12:00 UTC (n=15), (c) 18:00 UTC (n=20) and (d) 00:00 UTC (n=15), from radiosondes launched from the Savè ground site. Altitudes are above ground level (a.g.l.) (e) A time series of the RH at 1000 m. Dashed vertical lines indicate the four different monsoon phases, as defined by Knippertz et al. (2017). (NB: local time in Benin is UTC+1 h.)

Download

Using an average aerosol particle size of 200 nm (as measured by the SMPS) and assuming an aerosol composition of 52 % organic, 21 % sulfate, 14 % black carbon, 9 % ammonium and 4 % nitrate (see Fig. 2b), it was possible to estimate the average HGF of particles at each of the altitudes and times of day in Fig. 4 using the method described in Sect. 2.1 and 2.2. The value of κ for this composition was calculated using the method described in Sect. 2.1, assuming the composition to be the same across the size distribution. This produced a value of κ=0.22, which is typical for ambient aerosol (Pringle et al., 2010). These results are displayed in Fig. 5. The plots show the extent to which the diameter of an average particle would be expected to grow under the given conditions. The shape of the HGF profiles is governed by the RH statistics shown in Fig. 4. It can be seen that the sensitivity to changes in RH is more pronounced where the humidity is high. In drier regions, for example at low altitudes at 18:00 UTC, reasonably significant variation in RH produces very little change in particles' hygroscopic growth. This suggests that even thin layers of high RH are likely to have a disproportionately large impact on the radiative properties of the column. The particles are larger due to the high RH, and these larger particles scatter sunlight more effectively.

Ceilometer observations from the DACCIWA ground site in Savè during the campaign have been explored by Babić et al. (2018). This instrument uses a laser to determine atmospheric backscatter, which allows the determination of cloud base. A strong increase in backscatter was noted between 18:00 and 00:00 UTC, which is thought to be related to aerosol hygroscopic growth. The frequently high HGF layers shown here are consistent with this interpretation.

Figure 5The mean and standard deviation of estimated hygroscopic growth factors at (a) 06:00 UTC, (b) 12:00 UTC, (c) 18:00 UTC and (d) 00:00 UTC. Where the upper value cannot be seen is where the air became saturated.

Download

The volume of absorbed water at high humidities can make a significant difference to the composition and properties of aerosol particles. Figure 6a shows the percentage contribution of the compounds being considered here to the particle volume at different relative humidities. Above around 80 % RH, water contributes over half of the volume, and above around 95 % RH, water contributes more than 80 % to the volume of an individual particle, both growing the particle substantially and changing its optical properties. The decrease in aerosol refractive index as the particles grow is illustrated in Fig. 6b.

Figure 6(a) The fractional composition of particles at different relative humidities and (b) the effect of the additional water volume on the particle's refractive index. These values have been calculated using the ZSR mixing rule.

Download

3.2 Calculating aerosol optical depths

In order to establish the impact of humid layers on column optical properties, the sub-3 km aerosol optical depth (AOD) was calculated for each of the 63 unsaturated radiosonde profiles taken from the Savè ground site during the DACCIWA campaign. The steps of the approach used to establish the AOD are outlined below using a case study RH profile from 8 July at 10:57 UTC (local time = UTC+1 h). This approach was then applied to all RH profiles in the dataset:

Figure 7The average aerosol concentration profile from the area of interest. Circles and bars represent the median and the interquartile range, and the grey shading represents all data between the 10th and 90th percentiles. No flights went below 300 m in the dataset used here, so the open circle at 200 m is an estimation based on the median of data taken from flights further south. The histogram to the right shows the number of flights contributing data to each bin.

Download

1. Establish the average aerosol profile. The median aerosol mass concentration in the area of interest was calculated for each 100 m bin from 300 to 3000 m. Data were not collected from lower than this within this region, so an average data point from further south was used here to represent the 200 m bin. The resulting median profile can be seen in Fig. 7, alongside the interquartile range and 10th to 90th percentiles and the number of flights providing data in each bin. The median profile was used to calculate the AOD for all of the radiosonde profiles.

2. Establish an idealised aerosol size distribution at each altitude. Figure 3 shows the average aerosol size distribution at different altitudes, with two clear modes (Aitken and accumulation mode) being identified. In terms of aerosol hygroscopicity and changes to optical properties, the larger mode is the most significant as it contains the vast majority of the aerosol surface area and is of a size comparable with incoming radiation, making it the more optically active size range. The number distribution is dominated by the small, fresh Aitken mode aerosol, which shows a large degree of variation but contributes little to the overall optical properties of the aerosol; therefore, the accumulation mode was isolated here to calculate optical effects. It can be noted from Fig. 3 that there is little variation in the size and shape of the accumulation mode across different altitudes. The aerosol frequencies at different diameters generally fall within around 40 % of one another. This strongly suggests that a reasonably consistent accumulation mode can be expected, which will likely scale with the overall aerosol mass. In order to test the assumption that the Aitken mode does not contribute significantly to the AOD, the relative contribution of the average Aitken mode aerosol was calculated for a test case. The increase in AOD due to the addition of the Aitken mode aerosol was approximately 2.5 %. Given its limited contribution, only the accumulation mode aerosol was considered here.

   The thick black lines in Fig. 3 show log-normal distributions that have been fitted to the accumulation mode. The mode is at 200 nm. The shape and mode of the distribution were assumed to remain constant across all altitude bins and the amplitude was adjusted according to the median aerosol mass at a given altitude. This produced an idealised accumulation mode size distribution of the same shape, but differing amplitudes at each altitude.

3. Grow the wet aerosol size distribution. At each altitude bin, HGFs were used to “grow” the dry size distribution (see Zieger et al., 2013). An HGF was established and applied to each bin in the idealised dry size distribution, which allowed the shape of the new, wet size distribution to be calculated. An example can be seen in Fig. 8 for an idealised dry size distribution based on an aerosol mass of 10 µg m⁻³, assuming an RH of 90 %.

4. Calculate AOD. The chemical distribution, including water, of aerosol particles in each bin of the wet size distribution was used to calculate particles' refractive indices. Their extinction coefficients were then calculated at λ=525 nm using Mie code. The total extinction coefficient, measured per megametre (Mm⁻¹), at each altitude was determined by integrating these across the whole wet size distribution. This provides a quantification of the extent to which incoming radiation would be attenuated at each altitude. Integrating these with altitude gives the AOD of the column. In Fig. 9, the total AOD for the example radiosonde RH profile and the representative aerosol concentration are shown along with a profile of the dry and wet extinction coefficients at each altitude. The total AOD of the wet profile in this case is approximately 60 % larger than that of the dry profile.
   

   This example shows the disproportionate contribution towards AOD from layers of high humidity, such as that highlighted in grey in this example. Although the wet extinction coefficients are higher than the dry ones at every altitude, this effect is exaggerated in this humid layer.

Figure 8The dry and wet idealised (a) number and (b) volume size distributions based on an aerosol mass of 10 µg m⁻³ and an RH of 90 %.

Download

Figure 9The (a) relative humidity and (b) aerosol concentrations used to calculate AOD, and (c) the profile of dry and wet calculated extinction coefficients for the case study RH profile example on 8 July. The solid black line shows ground level at Savè. The dashed grey lines outline the humid layer with RH greater than 95 %. In the third panel, the darker filled shape shows the dry extinction coefficient in the column and the lighter filled shape shows the wet extinction coefficient.

Download

The approach outlined above has been used to calculate the AOD for all 63 unsaturated radiosonde RH profiles collected from Savè. Figure 10 shows a frequency plot of these calculated AODs, which are coloured according to the time of day the sonde was released (±1 h). These values are calculated purely based on variations in RH. The frequency distribution therefore quantifies an estimate of the variation in AOD due entirely to RH; it does not show a true estimate of AOD for each profile, which would require a larger number of varying parameters. This plot is shown alongside a frequency distribution of the AODs measured by sun photometer from the Savè ground site. The sun photometer dataset includes 319 measurements taken between 8 June and 29 July 2016 as part of the DACCIWA field campaign.

There is good general agreement between the AOD values calculated here and those collected by sun photometer from the Savè ground site over a similar period. The median AOD is 0.36 for both datasets. Values of AOD in the early morning have a larger spread, being reasonably well distributed between 0.3 and 0.7 in both cases, while AOD later in the evening in both cases was distributed more normally around the median. Both datasets have a high tail that is dominated by data from the mornings and the highest value is around 0.7 in the calculated dataset and 0.8 from the sun photometers. The assumption of a constant aerosol profile prevents the model from capturing the full spread that would result from variations in aerosol loadings. As a result, the sun photometer dataset has a larger interquartile range of 0.14, compared with 0.09 for the calculated dataset. The minimum value in the sun photometer dataset was a little less than 0.2, while that of the calculated dataset was closer to 0.3. This likely represents days with lower aerosol loadings, which are not captured by the model. The calculated dataset is based solely on observations in the lowest 3 km, and as such, any high-level aerosol layers are not taken into account. However, this does not appear to have significantly biased the calculated AODs.

It is important to note that the calculated values here are not expected to reproduce the real-world AOD for any given RH profile. A number of values are assumed to be constant here, including the shape and magnitude of the aerosol profile, the aerosol chemical composition, the shape of the accumulation mode size distribution and simplifications generally assumed in Mie calculations, such as that particles are internally mixed spheres. A realistic calculation would require variation in such quantities to be taken into account. This dataset, instead, gives an estimate of the variability in the AOD measurements that is due entirely to changes in RH. Despite this limitation, this approach is able to replicate a large part of the measured AOD frequency distribution; in particular, it is able to reproduce the high values of AOD observed during the mornings. While the inclusion of aerosol variation would be necessary in order to recreate the full spectrum of AODs, there are no observations of aerosol concentrations from anywhere across the region that are sufficiently high enough to be the sole cause of the enhancements in the ambient AOD observed by sun photometers. Although the omission of variations in aerosol properties has resulted in a narrower spread of calculated values compared with direct measurements, these results do show that the highly humid layers are the root cause of the very highest AODs being observed during the mornings.

Figure 10The frequency distributions of calculated AODs from radiosonde RHs, and AODs observed by sun photometers over Savé during the DACCIWA campaign. Bars shown here are stacked. Values have not been calculated for 00:00 UTC as this is during the night.

Download

3.3 The effect of high-humidity layers on AOD

In the example case study shown in Fig. 9, the humid layer between 500 and 900 m had a significant effect on the calculated column AOD. In Fig. 11, the relationship between highly humid layers and AOD is explored further. The maximum RH below 3 km was used as a simple parameter to quantify the presence of humid layers and the level of humidity within them for any given profile. The calculated AOD has been plotted here as a function of dry AOD ($f\left({\mathrm{AOD}}_{\mathrm{dry}}\right)={\mathrm{AOD}}_{\mathrm{wet}}/{\mathrm{AOD}}_{\mathrm{dry}}$) against the maximum RH for each of the 63 unsaturated RH profiles in Fig. 11. The f(AOD_dry) represents how much larger the calculated wet AOD is than the dry AOD, so a value of 1 is equal to the dry AOD. This figure clearly shows an increase in wet AOD above dry in all cases. However, when layers of extremely high RH are present, this effect becomes more pronounced. When humid layers with RH higher than 97 % are present, the wet AOD is more than 1.5 times larger than the dry AOD; with layers of RH greater than 98 %, the wet AOD can be almost 2 times greater or more than the dry. The use of f(AOD_dry) allows the potential for this value to be scaled according to the dry AOD from any aerosol profile.

A curve of best fit has been added to Fig. 11 using Igor Pro's iterative least-squares algorithm. The resulting formula can be used to calculate an estimate of the f(AOD_dry) from the dry AOD and the maximum RH in the profile (Eq. 5):

where RH is input as a value between 0 and 100. This is unlikely to be effective for values of RH_max significantly lower than those explored here. The factor of 1.25, which implies an f(AOD_dry) greater than the dry AOD when RH =0, is likely an artefact created by the use of only the maximum value of RH for each RH profile. This empirical relationship could be useful for models that have no detailed treatment of aerosol chemistry.

Figure 11(a) The relationship between calculated AOD and the maximum sub-3 km RH in a profile, and (b) the frequency distribution of maximum unsaturated sub-3 km RH values. Bars in (b) are stacked.

Download

The frequency distribution displayed in Fig. 11 shows that the RH reaches these high values a significant proportion of the time. Of the RH profiles gathered from Savé, 23 % included layers below 3 km that were unsaturated but contained layers of greater than 98 % humidity. This represents 37 % of the total number of unsaturated profiles.

A model-based study of aerosol liquid water content during the DACCIWA field campaign, carried out by Deetz et al. (2018a), found a mean dry AOD of 0.2 in southern West Africa, which increased to 0.7 when hygroscopic growth was taken into account. This mean dry aerosol is comparable with the dry AOD calculated here (0.25). The increase in AOD due to hygroscopic growth found here was slightly smaller: the median AOD calculated here was 0.36 (mean =0.39) and the median of the measured AOD from the sun photometers was 0.36 (mean =0.38). The study by Deetz et al. (2018a) considered a larger area, including regions further south and closer to the sea, which could explain the higher modelled hygroscopic growths in that study.

3.4 The effects of aerosol variability

Aerosol profiles have been assumed to be identical in all cases here. However, these loadings will in reality vary from day to day, which will have a resulting impact on the AOD. The extent of this impact can be estimated using the interquartile range of the aerosol measurements in the region, which are shown in Fig. 2. At all altitudes considered here, the interquartile range of the total aerosol loading was within 0.5 µg m⁻³ of 3.4 µg m⁻³. This is on average around a 27 % increase or decrease in the total aerosol population in less or more polluted circumstances, which would result in a similar change in the final calculated AOD. The median calculated AOD was 0.36; in this case, the less and more polluted circumstances would result in calculated AOD values of 0.26 and 0.46 respectively. This is a significant alteration in the final value and explains the broader range of AODs observed by the sun photometers.

The AOD scales directly with the aerosol loading, but scales exponentially with increases in RH. Therefore, the effects of a humid layer on a day that is more polluted than average could have an even larger effect than those shown here. The highest AOD values calculated in this study are around 0.7; this could scale to almost 0.9 if a day is among the 25 % most polluted. As anthropogenic emissions in the region increase, the number of days reaching these very high AOD values will increase. Furthermore, increased industrialisation can result in a higher proportion of inorganic, highly hygroscopic aerosol particles being produced (e.g. Morgan et al., 2009), which will exacerbate these effects.

These results suggest that the impact of highly humid layers on AOD, especially on days with high aerosol loadings, could be similar to that of low-level cloud. Capturing the presence of cloud in the region using satellites and other remote sensing techniques is known to be difficult in the southern West African region; the detection of these sub-saturated humid layers adds an additional challenge. Nonetheless, it is important for their presence to be taken into account if the radiation balance in the region is to be fully understood.

Due to the irregular nature of aircraft data and the focus of this study on the impacts of RH, a number of limitations to the scope of this study must be acknowledged. First, results shown here are based on observations of aerosol from the lowest 3 km of the atmosphere. In reality, some aerosol will exist at higher altitudes. For example, intrusions of biomass burning are known to become advected into this region regularly between 3 and 4 km at this time of year (Mari et al., 2008). Second, the focus on accumulation mode aerosol necessarily excludes the impact from other size modes from the calculations. Although very little coarse mode aerosol was observed by the Twin Otter aircraft during the DACCIWA campaign, it must be noted that any present aerosol in this mode would increase the region's AOD. The Twin Otter flew above 2750 m only once, but at this time, a higher concentration of sulfate aerosol was observed that this altitude. If this data point is representative of the altitude in general, then it is likely that this study has underestimated the HGF at the higher altitudes. All of these factors together suggest that the AOD calculations provided here are likely to represent a lower limit.