As expected, the major tracers of sea salt aerosols (Cl-, Na+, Mg2+ and K+) were sampled in high concentrations (up to 76, 53, 5.6 and 2.0 µg m-3, respectively) throughout the sampling periods. Their time variability, illustrated in Fig. 5 by the example of Na+, was very similar and characterised by a significant continuous background that could be represented by a 10-point moving average (that is, 90 h). The calculated mean background concentration was 10.1 ± 3.6 µg m-3. No seasonal cycle was evident due to the dominance of southerly and south-westerly winds transporting marine air masses onshore (Fig. 3).

The PMF sea salt component was represented by Na+, Cl-, Mg2+, K+, Ca2+ and SO42 (Fig. 6) and accounted for 74.7±1.9 % of the total aerosol mass (Fig. S3). Table 2 shows the mass ratios of Cl-, Mg2+, K+, Ca2+, F- and SO42- to Na+ for 2016 and 2017, calculated as the slopes of their linear regression lines and evaluated by the coefficient of determination (R2). This table also gives the slope of the linear regression lines for the PMF mineral dust component. The experimental values were compared with average ratios in seawater (Seinfeld and Pandis, 2006). The average Cl-/Na+ mass ratio was 1.4 ± 0.1 in 2016 and 1.3±0.1 in 2017 (also consistent for the PMF sea salt component), lower by 25 % than the value expected in seawater of 1.8. This difference has previously been reported in fresh sea salt in acidic marine environments (e.g. Zhang et al., 2010) and is attributed to Cl- depletion via reactions between NaCl and sulfuric and nitric acids. A very good correlation was observed between the ratios of Mg2+ (0.12 ± 0.01) and K+ (0.04 ± 0.01) with Na+ in this dataset and the value reported for seawater (Table 2) (Seinfeld and Pandis, 2006). Conversely, the linear correlation between Ca2+ and Na+ (not shown) was less pronounced (R2=0.61 and 0.42 in 2016 and 2017, respectively). The Ca2+ / Na+ mass ratio was systematically higher than in seawater (0.04), indicating the contribution of crustal calcium typical of the Namibian soils (see Sect. 4.2.2).

Using the average seawater ratio, the mean sea salt (ss) Ca2+ concentration was estimated as 470 ± 360 ng m-3 and 360 ± 210 ng m-3 for 2016 and 2017, respectively. The mean non-sea salt (nss) Ca2+ concentration was 420 ± 520 and 270 ± 400 ng m-3, respectively, for the two years, representing 47 % and 42 % of the mean measured Ca2+ concentrations. Similarly, for both 2016 and 2017, the ss and nss components of K+ were estimated as 367 ± 246 ng m-3 and 44 ± 54 ng m-3, respectively, accounting for 89 % and 11 % of the K+ mass. The PMF estimated that sea salt contributed 53.0 ± 1.6 % of the calcium and 75.1 ± 2.4 % of the K+ mass.

The mean F- / Na+ mass ratio measured at HBAO was 0.39 ± 0.29 in 2016 and 0.32 ± 0.29 in 2017 and was 0.19 ± 0.01 for the PMF sea salt component, enriched by 2 to 4 orders of magnitude to the average seawater composition (mass ratio 1.2×10-4; Table 2).

The PMF mineral dust component, composed of Si, Al, Fe, Ti, Ca2+, Mn, P, F- and V (Fig. 6), accounted for 15.7 ± 1.4 % of the total estimated mass. The time series of Al and nss-Ca2+ (Fig. 5) were analysed to investigate the temporal variability of airborne mineral dust at Henties Bay. The mean concentrations of mineral dust elements Al, Fe, Ti and Si were higher for night-time sampling between 21:00 and 06:00 UTC and lower in the day (09:00 to 18:00 UTC), in correspondence to easterly winds which were only observed at night and in the early morning (Fig. 4).

Differently from sea salt, the occurrence of mineral dust was not continuous, but episodic. Episodes of mineral dust corresponded to times when the concentrations of Al and nss-Ca2+ exceeded background values (modelled as the 10-point moving average) for a minimum of three consecutively sampled filters. Similar time variability was observed for elemental Fe, Si, Ti and P (not shown). Overall, 19 episodes of mineral dust were identified during the 2 years of sampling (Table S2).

The mean mass concentration of elemental Al was 556 ± 643 ng m-3 in 2016 and 446 ± 551 ng m-3 in 2017, while values peak as high as 4.7 µg m-3 (Table 1). To the best of our knowledge, no other measurements of Al are available in Namibia for comparison. Our arid sampling site is surrounded by loose sand, gravel plains (Matengu et al., 2019) and the deep Omaruru River valley directly north of the sampling site, which is also a recognised source of mineral dust to the offshore waters (Tlhalerwa et al., 2012). While mostly characterised by gravels, some clay-rich deposits are found around the river valley approximately 17 km north-east of HBAO (Matengu et al., 2019). The relatively low aluminium concentrations measured at HBAO suggest that these are not a major local source for the site. The nss-Ca2+ annual mean at HBAO (703 ± 644 ng m-3 in 2016 and 428 ± 437 ng m-3 in 2017) is similar to the concentrations (mean 425 ng m-3 and a maximum of 800 ng m-3) measured in central Namibia at Gobabeb, in the Namib Desert (23∘45′ S, 15∘03′ E; Annegarn et al., 1983). This is also the case for Fe, whose annual mean concentrations at HBAO (372 ± 480 ng m-3 in 2016 and 338 ± 433 ng m-3 in 2017) compare well with the average of 246 ng m-3 (Annegarn et al., 1983).

Table 3 shows the mass ratios for major components of mineral dust as well as some heavy metals (V and Ni). Overall, Si, Fe, and Ti showed very good correlations with Al, as expected for mineral dust (R2>0.9). The average mass ratio of Si / Al was 3.7 ± 1.0 in 2016 and 3.4 ± 0.8 in 2017, lower than the average values of 4 to 4.6 expected in global soils and crustal rock (Seinfeld and Pandis, 2006). This is attributed to the size fractionation during aeolian erosion of soils producing airborne dust. As a matter of fact, our average values are consistent with those obtained for particles less than 10 µm in diameter by Eltayeb et al. (1993) at Gobabeb. Our averages, generally higher than in mineral dust from northern Africa (Formenti et al., 2014), compare well with the value (3.4) reported by Caponi et al. (2017) for mineral dust aerosols generated in a laboratory experiment from a soil collected to the north-east of HBAO. The average Fe / Al ratio was 0.74 ± 0.19 in 2016 and 0.76 ± 0.18 in 2017 (0.8 ± 0.3 for the PMF solution), lower than the ratio of 1 reported by Eltayeb et al. (1993). The same is observed for the Ti / Al ratio, which was 0.07 ± 0.22 in 2016 and 0.06 ± 0.03 in 2017 (0.08 ± 0.01 in the PMF solution) but approximately 0.15 in Eltayeb et al. (1993).

The average nss-Ca2+ / Al ratio was 1.3 ± 0.7 in 2016 and 1.4±0.7 in 2017; however, for the strongest dust episodes (Al values higher than 1 µg m-3), the ratio tended to 1 (Fig. 7). This is in agreement with the specific mineralogy of Namibian soils that are rich in limestone and gypsum (Annegarn et al., 1983; Eltayeb et al., 1993). The PMF analysis attributed 40.5 ± 0.6 % of the total Ca2+ mass to the mineral dust component, of the same order of magnitude as obtained from the chemical apportionment (nss fraction representing 47 % of the total -Ca2+). The SO42- / Ca2+ mass ratio in the PMF mineral dust was 1.1±0.2, 3 to 4 times lower than the nss-SO42- / nss-Ca2+ obtained from chemical apportionment and about half the mass ratio for gypsum, which, however, coincided well with the mass ratio obtained when selecting the dust episodes only. The mean Fe / nss-Ca2+ ratio was 0.54 ± 0.23 in 2016 and 0.65 ± 0.23 in 2017, higher than the value of 0.11 ± 0.10 reported by Caponi et al. (2017), pointing to the diversity in soil mineralogy, even at relatively small spatial scales.

As for nss-Ca2+, values for nss-K+ / Al ratios (Fig. 7) were spread but ranged between 0.1 and 0.5 when Al concentrations exceeded 1 µg m-3. These values are in agreement with those for mineral dust sources in northern Africa (Formenti et al., 2014). The PMF K+ / Al mass ratio was 0.16 ± 0.01, in good agreement with the average nss-K+ / Al (0.13 ± 0.12) by chemical apportionment and half of that reported in the literature (0.25–0.45, Eltayeb et al., 1993).

The average phosphorus concentrations measured at HBAO were 11 ± 9 ng m-3 in 2016 and 14 ± 4 ng m-3 in 2017. Phosphorous was very well correlated with Al in 2016 (R2=0.92) and only moderately correlated in 2017 (R2=0.66). The P / Al mass ratio annual average was 0.03 ± 0.02 in 2016 and 0.05 ± 0.02 in 2017 (0.01 ± 0.01 in the PMF mineral dust). As was observed for the nss-Ca2+ / Al, the P / Al ratio tended to an asymptotic value of 0.02 when Al exceeded 1 µg m-3 (not shown). The PMF result is closer to that reported by Formenti et al. (2003a) for the outflow of Saharan dust to the North Atlantic Ocean (0.0070±0.0004).

The PMF identified two components characterised by heavy metals, a fugitive dust component (traced by V, Cd, Pb, Nd and Sr) and an industry component, characterised by As, Zn, Cu, Ni and Sr, representing 2.6 (±0.2 %) and 0.9 (±0.7 %) of the total estimated mass.

Vanadium and nickel are naturally occurring in mineral deposits in soils (Annegarn et al., 1983; Maier et al., 2013), but they are also known tracers of heavy-oil combustion, as reported in Becagli et al. (2017) and references therein. Their average concentrations at HBAO were 9 ± 5 ng m-3 (2016) and 7 ± 6 ng m-3 (2017) for V and 8 ± 7 ng m-3 (2016) and 7 ± 4 ng m-3 (2017) for Ni. The highest V concentrations corresponded to south-south-easterly winds, while high Ni concentrations were measured in the south-westerly wind sector (Fig. S4). The annual mean values of V and Ni at HBAO are an order of magnitude larger than measured over the open ocean by Chance et al. (2015), higher than those reported by Hedberg et al. (2005) at towns affected by copper smelters, and comparable to those measured by Isakson et al. (2001) at a Swedish harbour and by Becagli et al. (2017) in the central Mediterranean Sea downwind of a major shipping route.

Vanadium was well correlated with Al when Al exceeded 1 µg m-3 (R2 around 0.4), whereas no correlation between Ni and Al was observed (Fig. 7). Additionally, the correlation of V with Si, also used in the literature as a tracer of mineral dust, was evident while moderate (R2 around 0.4), and no correlation was found for Ni. This differs from what was reported by Becagli et al. (2017), who found that neither V nor Ni was correlated with Si. In our dataset and the PMF mineral dust component (Sect. 4.2.2), both V / Si and Ni / Si ratios were enriched by a factor of 10 or more to reference values for the upper continental crust (3.1×10-4 and 1.5×10-4 for V / Si and Ni / Si, respectively; Henderson and Henderson, 2009). The V / Ni mass ratio was 1.7 ± 1.1 for 2016 and 1.3±1.3 in 2017, lower than reported by Lyyränen et al. (1999) and Corbin et al. (2018) for heavy fuel oil in diesel engines and by Becagli et al. (2017) and Viana et al. (2009) in the Mediterranean basin ambient air (2.8–2.9 and 4–5, respectively).

All these elements, and furthermore their poor correlation (R2 around 0.3), suggest that V and Ni do not necessarily have the same sources. Mining activities, likely in the Otavi mountain area (Boni et al., 2007), should account for the high concentrations of V, with additional contributions from heavy-oil combustion, where V is present as an impurity (Isakson et al., 2001, and references therein; Vouk and Piver, 1983). By contrast, combustion of heavy oils seems to be the primary source of Ni.

This hypothesis is supported by the PMF analysis. The PMF apportionment of V and Ni concentrations (Fig. S5) clearly distinguishes the relative source contributions and preferentially associates V with the mineral dust and fugitive dust components but Ni with the industry component.

Moderate to good correlations of V and Ni with Zn (R2 of 0.42 and 0.55, respectively), Cu (0.55 and 0.73) and Pb (0.56 and 0.69) were observed in the dataset. Zn and Pb are found as impurities in bulk fuels for ships (Isakson et al., 2001) and also from copper smelting, as reported in central Chile (Hedberg et al., 2005) and urban air in the United States of America (Ramadan et al., 2000). The mean concentration of Zn at HBAO (11 ± 9 ng m-3) was about 2 orders of magnitude higher than over the south-eastern Atlantic Ocean (Chance et al., 2015) and in air over the arid landscapes (Annegarn et al., 1983). Likewise, the mean Pb concentration (75±89 ng m-3) was 3 orders of magnitude higher than reported by Chance et al. (2015) for soluble Pb and comparable to values measured in the western Mediterranean by Denjean et al. (2016). The PMF separates the largest fractions of Zn and Pb into the industry and fugitive dust components, respectively. Although some of these heavy metals may be sourced from the commercial shipping route offshore, the mass ratios for tracer elements were not in agreement with our results, and so we cannot conclusively name shipping heavy-oil combustion as the source of these heavy metals.

Average concentrations of Cu at HBAO were 8 ± 6 ng m-3, an order of magnitude higher than measured in windblown dust by Annegarn et al. (1983) in the central Namib but 2 orders of magnitude smaller than the average measured by Lee et al. (1999) in highly polluted Hong Kong (125.1 ng m-3). Ettler et al. (2011) showed that copper ore mining and smelting operations in the Zambian copper belt are a significant source of potentially bioavailable copper that, unlike phosphorus, has been found to inhibit plankton growth in laboratory studies (Paytan et al., 2009) and over the western Mediterranean (Jordi et al., 2012). Similar contamination of topsoil was found by Kříbek et al. (2018) at operations in the Tsumeb mining district, Namibia (19∘14′ S, 17∘43′ E). Average Cu concentrations were comparable to values of 4.9 ± 11.5 ng m-3 reported for a town closer to smelters in Chile and an order of magnitude smaller than in the urban environment of the capital city of Santiago (77.5 ± 78.2 ng m-3; Hedberg et al., 2005). The Cu / Ni ratio (1.24 ± 0.20) in the PMF fugitive dust component was about half that reported for soil samples polluted by copper mine tailings from the Gruben River valley (2.03 ± 2.30, Taylor and Kesterton, 2002).

The mean mass concentration of Cd was 1502 ± 1458 ng m-3 in 2016 and 219 ± 163 ng m-3 in 2017. The difference is mainly due to high concentrations in October of 2016 which coincided with high concentrations in all other heavy metals, except for As. Cd concentrations in 2016 were less than that reported for airborne road dust (7.4 ± 7.8 µg m-3), and our 2017 concentrations were of the order of that measured in ambient air (0.14 ± 0.04 µg m-3) in the seaside city of Khobar, Saudi Arabia (El-Sergany and El-Sharkawy, 2011). The Cd / Pb ratio of 9.96 ± 0.21 for the PMF fugitive dust component was slightly higher than 7.14 ± 4.26 in the ambient air of the coastal desert environment in Khobar (El-Sergany and El-Sharkawy, 2011). The correlation of Pb, Nd, Sr in the fugitive dust component may indicate contributions of non-micaceous kimberlites from a variety of source regions across southern Africa (Smith, 1983). The Sr / Nd ratio for the fugitive dust component (3.58) was close to the 3.35 reported for kimberlites at Uintjiesberg in the Northern Cape of South Africa.

One of the striking features of Table 1 is the high mean concentration of F- measured at HBAO (4.3 ± 4.0 µg m-3 in 2016 and 2.8 ± 2.5 µg m-3 in 2017), with peak values as high as 25 µg m-3. Those annual mean concentrations were comparable to the mean 24 h fluoride concentrations measured between 1985 and 1990 over the South African Highveld by Scheifinger and Held (1997). The measured concentrations at HBAO were also comparable to those of heavily polluted areas in China (Feng et al., 2003) and significantly higher than reported for Europe, even in the polluted Venice lagoon (Prodi et al., 2009) or in areas nearby ceramic and glass factories (Calastrini et al., 1998). The peak values at HBAO were significantly higher than maxima reported by these authors and ranging between 1.4 and 2.9 µg m-3. The highest F- concentrations were associated with southerly to easterly winds, that is, from the subcontinent (not shown). The very good correlation of F- with nss-Ca2+, shown in Fig. S6 (R2 equal to 0.76 in 2016 and to 0.84 in 2017), yielded a mean mass ratio of 6.4 and 5.8, respectively, much higher than reported in groundwater, aerosols or precipitation in polluted environments (Feng et al., 2003; Prodi et al., 2009).

The strong relationship with nss-Ca2+ (and a posteriori with Ca2+) drove the PMF apportionment (Fig. S7), which attributed approximately 94 % of the F- mass concentrations to the sea salt and mineral dust components (55.1 ± 1.9 % and 38.8 ± 1.1 %, respectively) and the remaining 6 % to fugitive dust (2.3 ± 0.5 %) and industry (3.8 ± 1.0 %). Possible sources are the emission of fugitive dust during fluorspar mining of carbonatite-related fluorspar deposits at the Okorusu Mine (20∘3′ S, 16∘44′ E) but very likely also the periodic surface mining occurring approximately 20 km south of HBAO to provide gravel for the construction of a major road between Swakopmund and Henties Bay which started late in 2015 (Andreas Namwoonde, personal communication, 2017). The evaporation of fluoride-rich water, leached into groundwater (Wanke et al., 2015, 2017) from fluoride-rich mineral deposits and soils throughout the region and in the coastal waters (Compton and Bergh, 2016; Mänd et al., 2018), would also increase atmospheric F- concentrations. In an analysis of borehole water in Namibia, roughly 80 % of those sites surveyed were deemed unsafe to drink as a direct result of high fluoride concentrations (Christelis and Struckmeier, 2011).

The annual mean of the arsenic concentrations at HBAO was 22 ± 16 ng m-3 in 2016 and 239 ± 344 ng m-3 in 2017. The mean for 2017 is skewed due to two sampling weeks with very high concentrations in the order of those measured in rural and urban-industrial areas affected by mining and smelting emission sources (Hedberg et al., 2005; Šerbula et al., 2010).

The PMF analysis exclusively associated As the industry component along with large fractions of the Zn, Cu, Ni, Sr and Co. Known sources of atmospheric arsenic are biomass burning, heavy-oil combustion and non-ferrous metal smelting operations (Ahoulé et al., 2015; Gomez-Caminero et al., 2001). A possible local source could be the Tsumeb smelter to the north-east of HBAO (KPMG, 2014).

The PMF As / Zn, As / Pb and Zn / Pb ratios were 9.0 ± 0.3, 6.4 ± 0.8 and 0.7 ± 0.1, in good agreement with those reported by Hedberg et al. (2005) for a copper smelter plume in Chile (7.7, 4.5 and 0.6, respectively). This is in good agreement with the fact that no correlations between As and Al or nss-Ca2+ were found, ruling out any major contribution of inorganic arsenic in geologic formations released from mining operations or evaporated from soil and groundwater (Gomez-Caminero et al., 2001). Likewise, no discernible correlation between As and MSA was found, suggesting only a minor release of arsenic by marine algae and plankton (Sanders and Windom, 1980; Shibata et al., 1996).

The PMF ammonium neutralised (Fig. 6) comprised secondary species such as by SO42-, NH4+, MSA, oxalate, and nitrate, which accounted for 6.1 ± 0.7 % of the estimated aerosol mass.

The annual mean sulfate concentration measured at HBAO was 4.1 ± 2.6 µg m-3 in 2016 and 3.4 ± 1.4 µg m-3 in 2017 (Table 4), higher than previously measured over the southern Atlantic and Pacific oceans (Zhang et al., 2010) and comparable to springtime measurements in the Venice lagoon (Prodi et al., 2009). As already discussed in Formenti et al. (2019), the highest concentrations were measured in spring and autumn, while minima occurred between May and August. SO42- and Na+ showed good correlation (R2=0.92 in 2016 and 0.83 in 2017, Table 2). However, their annual mass ratios (0.36±0.14 and 0.42 ± 0.23 in 2016 and 2017, respectively) were higher than the expected mass ratio in seawater (0.25; Seinfeld and Pandis, 2006), which was used as a nominal reference to apportion SO42- into its ss and nss fractions. As a result, up to 57 % of the measured SO42- mass concentration in the PM10 fraction was attributed to sea salt aerosols, while the nss component was of the order of 43 %. The PMF estimated that the sea salt component contributed 66.6 ± 0.4 % of the total sulfate mass. This is in agreement with previous observations in the South Atlantic Ocean (Andreae et al., 1995; Zhang et al., 2010; Zorn et al., 2008). By contrast, at the remote Brand se Baai site along the Atlantic coast of South Africa (31.5∘ S, 18∘ E), Formenti et al. (1999) reported that sea salt accounted for about 92 % of the total measured elemental sulfur concentrations.

The MSA concentrations measured at the site ranged between 10 and 230 ng m-3 (Table 1). The mean annual concentration was 63 ± 39 ng m-3, 3 times higher than the mean value of 20 ± 20 ng m-3 (6.2 ± 4.2 ppt) reported by Andreae et al. (1995) over the open ocean along 19∘ S and lower than in the south-eastern Atlantic Ocean (Zhang et al., 2010; Table 4). As already described in Formenti et al. (2019), the MSA concentrations were higher in the austral summer and spring and lower in the austral winter. DMS is more efficiently oxidised in warmer conditions (Ayers et al., 1986; Huang et al., 2017), which explains the higher daytime mean concentrations of marine biogenic products (MSA and nss-SO42-) and lower means at night and in the winter. Springtime averages for MSA were in the range of that measured by Huang et al. (2017) during a springtime cruise over the South Atlantic and by Prodi et al. (2009) in the Venice lagoon (Table 4). The mismatch of seasonality with respect to that of the phytoplankton blooms (Louw et al., 2016) has already been discussed by Formenti et al. (2019) and attributed to the spread of blooms in the BUS region depending on local conditions.

The MSA / nss-SO42- ratio (Fig. 8) displayed a large range of values (0.01 to 0.12), consistent with that reported in the literature at various geographical locations, especially in the Southern Hemisphere (Table 4). The MSA / SO42- mass ratio for the PMF component (0.04 ± 0.01) was in agreement with the MSA / nss-SO42- from the chemical apportionment reported in Table 4. The strong seasonal dependence of MSA / nss-SO42- is in agreement with that identified by Ayers et al. (1986) for marine biogenic sulfur in the Southern Hemisphere and suggests that the highest concentrations of nss-SO42- in the PM10 (nss-SO42- larger than 2 µg m-3) are not necessarily associated with marine biogenic emissions. From measurements at the desert station of Gobabeb, in the Namib Desert, Annegarn et al. (1983) found that only the fine mode of the bimodal distribution of sulfur aerosols, that is, that bearing the lower mass concentrations, would be due to the oxidation of sulfur-containing gaseous emissions during the marine phytoplankton life cycle.

Figure 8 illustrates the NH4+ / nss-SO42- mass ratio as a function of nss-SO42- mass concentrations. In both 2016 and 2017, the NH4+ / nss-SO42- mass ratios were less variable than for MSA / nss-SO42-. The annual mean NH4+ / nss-SO42- were 0.13 ± 0.10 in 2016, 0.14±0.08 in 2017, and 0.15 ± 0.01 in 2017. These values are consistent with the mass ratio of 0.18 corresponding to ammonium bisulfate ((NH4)HSO4). Although some losses of NH4+ due to conservation on site and transport to the laboratory in France cannot be excluded, the measured ratios are consistent with previous investigations in remote marine environments reported in Table 4, including offshore southern Africa (Andreae et al., 1995; Quinn et al., 1998).

The average NO3- / nss-SO42- ratio at HBAO was of the order of 0.14, significantly smaller than reported by Zhang et al. (2010) over the south-eastern Atlantic. Poor correlation between nss-SO42- and nss-Ca2+ (not shown) suggests that very little of the sulfate is present as CaSO4, either formed by heterogeneous deposition of SO2 on calcite mineral particles or liberated from the soils as mineral gypsum (Annegarn et al., 1983).

Finally, the mean annual concentration of oxalate at HBAO was 72 ± 80 ng m-3 in 2016 and 141 ± 50 ng m-3 in 2017. Values at HBAO are consistent with those reported by Zhang et al. (2010) over the south-eastern Atlantic (200 ± 140 ng m-3). Oxalate aerosols in the atmosphere are due to marine biogenic activity and anthropogenic emissions including heavy-oil combustion and biomass burning (Gillett et al., 2007, and references therein). They are also formed by in-cloud processes and oxidation of gaseous precursors followed by condensation (Baboukas et al., 2000). The moderate correlation with NO3-, nss-SO42-, and nss-K+, particularly in 2017, could suggest a common origin and possible influence of occasional biomass burning.