Interhemispheric transport of viable fungi and bacteria from Africa to the Caribbean with soil dust

Transcription

1 Aerobiologia (2005) 21:1 19 Ó Springer 2005 DOI /s Interhemispheric transport of viable fungi and bacteria from Africa to the Caribbean with soil dust Joseph M. Prospero 1, *, Edmund Blades 2, George Mathison 3 & Raana Naidu 4,5 1 Cooperative Institute for Marine and Atmospheric Studies, Rosenstiel School of Marine and Atmospheric Science, University of Miami, 4600 Rickenbacker Causeway, Miami FL 33149, USA; 2 Faculty of Pure and Applied Sciences, University of the West Indies and Queen Elizabeth Hospital, Barbados, West Indies; 3 Faculty of Pure and Applied Sciences, University of the West Indies, Barbados, West Indies; 4 Faculty of Medical Sciences, University of the West Indies and Queen Elizabeth Hospital, Barbados, West Indies; 5 Present address: Greenville Hospital System, 701 Grove Rd, Greenville SC (*Author for correspondence, jprospero@rsmas.miami.edu; Fax: ) (Received 4 May 2004; accepted in final form 22 October 2004) Key words: African dust, bacteria, fungi, intercontinental transport, microorganisms Abstract Daily aerosol samples collected in trade winds at Barbados, West Indies, throughout yielded significant concentrations of viable (culture-forming) bacteria and fungi only when African dust was present. Air masses from the North Atlantic, North America, and Europe yielded no cultivable organisms. The strong association of cultivable organisms with African dust suggests various factors that might be relevant to viability. Although we did not specifically look for pathogens, these same mechanisms could protect them as well. Our results suggest that arid regions could be an important source for the long-range transport of viable microorganisms. The transport of microorganisms to Barbados follows a clear meteorological and seasonal pattern, which suggests that it should be possible to model the transport process and to predict events. Microorganism and dust concentrations were unusually great in 1997, possibly in response to the strong El Nin o. This suggests that the long-range transport of microorganisms might be particularly responsive to climate variability in general. 1. Introduction Winds serve as a mechanism that enables the rapid transport of microorganisms (MOs) among widely dispersed habitats (Isard and Gage, 2001). These include organisms pathogenic to humans, animals, and plants (Brown and Hovmøller, 2002). It is well known that plant pathogens can be transported by winds over hundreds of kilometres, for example, the periodic transport of Tobacco Blue Mold (Peronospora tabacina Adam) from Cuba to the southeastern United States (Davis and Monahan, 1991). There is, however, only anecdotal indirect evidence for the long-range transport (LRT) of viable MOs on intercontinental scales. In the review by Brown and Hovmøller (2002), the attribution to over-ocean wind transport of MOs is based solely on the existence of generally favorable synoptic meteorological conditions, not on any specific measurements of MOs present in hypothesized transporting winds. For example, sugarcane rust in the Dominican Republic is attributed to the transport of urediospores from Cameroon (Purdy et al., 1985), based on the relative timing of rust outbreaks in those locations and the prevalence of wind trajectories that could

2 2 conceivably effect such transport. The one unambiguous clearly documented case of live organism transport from Africa to the Caribbean occurred in 1988, when swarms of African desert locusts (Schistocerca gregaria) reached Trinidad and Barbados (Ritchie and Pedgley, 1989; Rosenberg and Burt, 1999). We are aware of no long-term systematic studies of the LRT of MOs over the oceans. A few studies have been carried out over the oceans and in remote (primarily high latitude) regions but the data are sporadic and much of it is in the older literature. Lighthart and Stetzenbach (1994) summarize this literature and conclude that although it has been shown that bacteria and fungi can be transported in the air for long distances, evidence suggests that most survive over relatively short times. They also note that the genera/species distributions change with distance from the source but they do not comment on any systematic behavior that would provide insights on the factors that affect the viability of the MOs. Our study is based on Barbados (13.18 N, W), the eastern-most island in the Caribbean (Figure 1), where we have made aerosol measurements continuously since During much of the year the trade winds carry large quantities of African mineral dust to Barbados (Savoie et al., 1989; Li et al., 1996; Prospero 1996; Prospero and Lamb 2003). Winds subsequently carry dust into the Caribbean and to Florida (Prospero, 1999; Prospero et al., 2001), to the eastern United States (Perry et al., 1997) and to Bermuda (Arimoto et al., 1992, 1995). The primary purpose of this study is to characterize the day-to-day variability of common airborne MOs and to place the variability in the context of the large-scale synoptic meteorology of the tropical Atlantic. To this end we collected daily aerosol filters and cultured them for bacteria and fungi using standard techniques. Concurrently we measured an array of inorganic aerosol species: mineral dust from soils; non-sea-salt (nss) SO 4 = i.e., SO 4 = from sources other than the salts in ocean-water spray droplets, primarily pollution sources and the oxidation of dimethyl sulfide emitted by marine phytoplankton (Davis et al., 1999); NO 3 ), derived from both natural and pollution sources (Holland et al., 1999); sea-salt from ocean spray droplets (Savoie et al., 2002). Because of our long history of studies on Barbados we have a good understanding of the factors that control aerosol composition and concentration in the North Atlantic trade winds. By combining measurements of conventional inorganic aerosols, including mineral dust, with those of bacteria and fungi, we would expect to gain a better understanding of the factors affecting the large scale transport of viable MOs. Because we collect daily samples we can more readily relate our measurements to specific meteorological situations and to specific back-trajectories. In this study we examine and compare the daily temporal record of aerosol and microorganism concentrations over the period Methods 2.1. Sampling techniques Figure 1. Map of Barbados with inset map of the tropical Atlantic. The Barbados map shows the location of the University of Miami sampling station at East Point immediately to the north of Kitridge Point. The X shows the location of the sampling site at the University of the West Indies, Cave Hill; the star sign (q), the location of the inland sampling site on the western side of the island. 1 km from the east coast. Inset: Map of the North Atlantic showing the location of Barbados along with other aerosol sampling stations occupied at various times in the University of Miami aerosol program. Sampling is carried out at Ragged Point, Barbados, (Figure 1) at the top of a 17 m high walk-up tower standing on a 30 m high bluff located immediately on the easternmost coast of the island. Daily filter samples are collected at a nominal 10 l min )1 (i.e., about 14 m 3 day )1 ) using sterile microbiological filters (cellulose nitrate, 47 mm in diameter, 0.2 lm pore size). We paid considerable attention to the location of the sampling site and sampling

3 3 protocols so as to minimize the possibility of impacts from local sources. We later show that air sampled a short distance inland from the coast of Barbados can yield cultivable MO concentrations that are about times higher than those obtained at our tower site. Among other precautions we have used smoke tests to show that the top of the tower lies above the turbulent air layer generated by the wind flowing up the face of the bluff and over the top of the bluff. The pump is controlled by a wind sensor via a computer so as to sample only over-ocean winds in the sector 335 through North to 130 at wind speeds greater than 1 m sec )1. Within the sampling sector the closest land is the Cape Verde Islands and the coast of Africa, 3800 and 4600 km, respectively, to the east. Sampling is continuous during the 24-hour period so long as the wind sector condition is met (on average, about 95% of the time) and it is not raining. Two filters are deployed at the top of the tower each day, one serving as a blank, which is processed in the same way as the sample. Working onsite in a HEPA filtered clean bench, filters are placed in sterile Petri dishes and transported to the microbiology laboratory of the Queen Elizabeth Hospital where the School of Clinical Medicine and Research, University of the West Indies, is located. All subsequent manipulations were carried out in the Microbiology Department of the Pathology Laboratory using procedures normally used in the hospital s microbiological analyses Culturing techniques and identifications Bioaerosol measurements are very sensitive to collection and culturing techniques (Burge, 1995; Lacey and Venette, 1995; Griffiths et al., 1999, 2001). In this study our objective was to characterize the temporal variability of bacteria and fungi that are commonly found over land. To facilitate comparisons with previous work we used techniques that are frequently used in such studies. Filters were cut in half with a sterile scalpel. One section was placed (sample side up) on blood agar medium for bacterial growth. Blood-based nutrients are non-selective media which are widely used for broad-spectrum studies (Lacey and Venette, 1995). The other half filter was placed on Sabouraud s medium for fungi. The cultures were initially incubated at 37 C for 48 hours and then at 30 C for up to 2 weeks. The number of colonies was counted and the results reported as the number of colony-was forming units (CFU) per cubic meter of air (CFU m )3 ). The resulting cultures were subcultured onto similar media. With rare exceptions, blanks yielded no cultures. Standard light microscopy and phase contrast microscopy were used to examine fungal cultures and identify spores produced in culture. Phenotypic identifications were based on the macroscopic and microscopic morphology of the cultures and spores, respectively. When a clear identification could not be made, the sample was sent to the University of Texas Health Science Center, San Antonio (M.G. Rinaldi), where reference collections are maintained. Bacterial endospores are formed by certain genera. Under favorable conditions these spores germinate to produce vegetative cells. Unstained endospores are not readily detectable under light microscopy, especially with the high loading of dust on the surface of the filters. Identification is based primarily on morphology, gram stain, and spore stain of the cultures. Staining for spores was performed on 1-week-old cultures (Murray et al., 1999). Bacteria were typically gram positive (stain purple) or gram variable (some stain purple and some stain pinkish). Some gram positive organisms, especially Bacillus species, lose their gram positive properties with age and therefore tend to stain pink. On a limited number of samples we searched for anaerobic species (e.g., Clostridia). We cultured filters anaerobically, using Oxoid Anerogen gas generating kits with indicator; none were found. Because there is some debate about the relative merits of rich culture media, such as Sabouraud s, and weaker media (Lacey and Venette, 1995; Burge, 1995), we ran tests to compare Sabouraud s (the one normally used in our program) with R2A, a weaker nutrient commonly used in the culture of stressed organisms (Burge, 1995; Muilenberg, 1995). Cultures of daily filters collected over 17 days of concurrent sampling yielded similar results with both media. Linear regression of Sabouraud vs. R2A yields: b(0) ¼ 0.038, b(1) ¼ 1.19, r 2 ¼ One sample yielded concentrations (CFU) about ten times greater that the highest measured in the remaining 16 samples; eliminating this one outlier, the regression yields comparable values: b(0) ¼ 0.157, b(1) ¼ 0.88,

4 4 r 2 ¼ We conclude that there is no substantial difference between these two media with regard to the culturing of organisms associated with African dust. We also ran tests for the effects of varying sampling duration to see if viability is affected by long exposure to air streaming through the filter and to high loadings of sea salt, factors that can affect MO viability (Lighthart, 2000). We used three parallel samplers. One ran continuously (as long as the wind condition was satisfied) for a nominal 24 hours, the normal sampling protocol. The other two samplers were run (also wind-sector controlled) on skip timers with one running intermittently for six equally spaced 1-hour periods (total 6 hours) over a nominal day and the other running intermittently for six quarter-hour periods (total 1.5 hours). Filter halves were cultured on both Sabouraud s and R2A media. Over 16 days of daily sampling we found no substantial differences between the two nutrients regardless of sampling time. The 24-hour samples yielded a mean of 0.51 CFU/m 3 (standard deviation ¼ 0.38) with Sabouraud s and 0.40 CFU/m 3 (0.38) with R2A. Yields for 6-hour samples: Sabouraud s, 0.47 CFU/m 3 (0.71); R2A, 0.89 CFU/m 3 (1.33). For 1.5-hour samples: Sabouraud s, 0.96 CFU/m 3 (1.48), R2A, 0.22 CFU/m 3 (0.86). As expected the shorter sampling times yielded poorer statistics because of the small volumes and low colony counts Inorganic aerosol sampling In addition to the biological samples, we concurrently collected daily aerosol samples using Whatman-41 filters and the same wind-sector controller. Filters were returned to Miami where the soluble fraction was extracted and analyzed for SO 4 =,NO 3 ), and Na + (Savoie et al., 2002). Sulfate and NO 3 ) concentrations were measured with 1r uncertainties of ±5% using suppressed ion chromatography and Na + with a 1r uncertainty of ±2% by flame atomic absorption spectrophotometry. Sodium was used to calculate sea-salt aerosol concentrations, multiplying by 3.252, the Na + /total-salts ratio in sea water. The concentration of SO 4 = from sources other than the dissolved salts in seawater (nss-so 4 = ) was calculated as total SO 4 = minus the Na + concentration times , the SO 4 = /Na + mass ratio in bulk seawater. The absolute 1r analytical uncertainties for measurements of nss-so 4 = were usually less than 0.1 lg m )3 ; uncertainties due to blank corrections were typically about an order of magnitude lower. The extracted filters were then placed in a muffle furnace for about 14 hours (overnight) at 550 C; the ash residue weight (less filter blank) is assumed to be mineral dust. The standard error in the mineral aerosol concentration is ±10% for concentrations greater than about 1 lg m )3 ; below 1 lg m 3 the standard error is essentially constant at ±0.1 lg m )3. The parallel analyses of samples for Al yields an Al content of about 8%, a value consistent with average crustal composition (Prospero, 1999). 3. Results 3.1. Temporal variability The cultivable fungi and bacteria were primarily spore-formers (Table 1). Concentrations varied over a wide range (Figure 2), from 0 to about 20 CFU m )3. Concentrations were mostly zero during the winter; they increased sporadically during the spring, remained relatively high through the summer, and then showed sporadic behavior once again in the fall. There was a close match in the temporal spacing of the peaks of fungi and bacteria but there were often substantial differences in their relative concentrations. For exampled in the spring and summer of 1996 fungi tend, to be somewhat greater than bacteria whereas bacteria strongly dominated in the fall. In 1997 bacteria dominated in the spring but there are specific events where fungi concentrations are much higher. Indeed, scatter plots of the concentrations of bacteria against fungi (not shown) yield a very broad distribution with no apparent correlation. There are also substantial differences between the years. In particular, during February March and most notably in June 1997 the concentrations of both bacteria and fungi were markedly higher than in In fact concentrations in June 1997 were substantially higher than any other period in the record. The June 1997 anomaly stands out clearly in Figure 2. Figure 3 shows the daily record of fungal and bacterial concentrations in displayed along with the concentrations of mineral dust, nss-so 4 =, and sea-salt. Note that the monthly means in many months (e.g., Fall

5 5 Table 1. Comparison of fungi in Barbados trade winds with African and Middle East regions a This work Barbados Dransfield (1966) b Northern Nigeria Davies (1969) c Kuwait Halwagy (1989) d Kuwait Abdalla (1988) e Khartoum, Sudan Al-Subai (2002) f Qatar Al-Suwaine et al. (1999) g Ridyadh % Air % Air % Soil % Air % Air % Air % Air % Air Mycelia sterila Arthrinium Periconium 12.0 Black/grey 6.9 Brown/tan 3.1 Penicillium Curvularia Cladosporium Aspergillus (total) A. niger A. fumigatus A. clavatus 0.08 A. terreus A. flavus A. nidulans 8.1 A. sydowi 0.06 Neurospora Pink/red 0.03 Alternaria Epicoccum Fusarium Nigrospora Pullularia Ulocladium Ustilago a We report all cultures identified in Barbados tower samples, the corresponding species found in the literature cited here as well as major species found in one or more of the cited literature but not found in Barbados samples. b Dransfeld (1966): Air deposition by settling to Pietri dish; culture on 2% Difco malt agar. Soil: Czapek-Dox agar. c Davies (1969): Spore trap, microscope. d Halwagy (1989): Spore trap, microscope. e Abdalla (1988): Settling from air on potato dextrose agar. Focused mostly on Aspergillus. Listing here for June. f Al-Subai (2002): Settling onto glucose agar. g Al-Suwaine et al. (1999): Suction air sampling onto agar plates. Average of two sites.

6 6 Figure 2. The concentration of cultivable fungi (right axis) and bacteria (left axis) in the trade winds at Barbados, Units: number of colony-forming units (CFU) of fungi and bacteria per cubic meter of air sampled through the filter (CFU m 3 ). The scale on the right-hand axis is offset for clarity. Figure 3. The daily concentration of cultivable fungi and bacteria in Barbados trade winds during 1996 and 1997 in comparison to various aerosol constituents: (a) bacteria and mineral dust; (b) fungi and dust; (c) fungi and nss-sulfate (nss-so 4 = ); (d) fungi and sea salt. Units: fungi and bacteria, CFU m )3 ; aerosols, lg m )3. In each panel the MO concentration is shown in blue and the comparison aerosol (dust, nss-so 4 =, sea-salt) in red.

7 7 1996) are strongly driven by a few samples with relatively high concentrations. This is especially true for bacteria. Fungal concentrations do not seem to be quite so variable. Figure 3 shows the daily record of fungal and bacterial concentrations in displayed along with those for mineral dust, nss-so = 4, and sea-salt. Fungi and bacteria show a seasonal cycle that is very similar to that of dust (Figure 3a, b): extremely low concentrations in winter, high in summer. Dust concentrations were much higher in February March 1997 and the summer and early fall of 1997 compared to 1996, similar to the previously noted differences in the interannual concentrations of fungi and bacteria. Furthermore there is a close match in the temporal spacing of individual pulses of dust and the peaks of fungi and bacteria. The close correspondence between dust pulses and those of fungi and bacteria is especially notable in June Although the timing of the peaks matches well, the concentrations of fungi and dust were not correlated; scatter plots (not shown) yield very broad distributions. Scatter plots of bacteria against dust against (not shown) are similar. In contrast to dust, there is no coherence between the peak patterns of bacteria and fungi and those of nss-so = 4 and sea-salt (Figure 3c, d). Note in particular the winter and spring periods (1 January to late April 1996; December 1996 to late May 1997) when fungi and bacteria concentrations were extremely low (usually zero) along with dust except for occasional brief pulses. During this time sea-salt concentrations were quite high because of strong winter winds while those of nss-so = 4 were often substantial. Ocean spray droplets can carry large concentrations of bacteria and fungi but culturing these would require specific nutrient media. To examine more carefully the relationship between dust concentrations and those of bacteria and fungi we show data (Figure 4) for three major dusty periods. In early 1996 (Figure 4a) fungi concentrations increased dramatically from background levels with the first pulse of dust on 31 March. Bacteria also increased but only moderately. Fungi concentrations dropped sharply on 13 April with the end of the extended 2-week dusty period. Throughout this dust event fungi concentrations were always much greater than those of bacteria. After 13 April fungi and bacteria remained at background levels until 5 May when the next pulse of dust arrived. After 5 May fungi concentrations were generally greater than bacteria although the difference was not so great as in the April event; on some days bacteria dominated fungi. This example more clearly shows a point made previously: that we only see MOs during dust events but that within dust events MOs are not correlated with dust. Indeed, some dust events yield very low cultivable MOs. Note in particular during 5 7 April when dust concentrations were high, fungi dropped to low levels. Similar examples are noted elsewhere in this time series. In early 1997 (Figure 4b), we see once again that MO concentrations are at background levels (i.e., essentially none) until the first dusty period. Although we obtained no MO samples during the very beginning of this dust event which started on 6 7 February, it is clear that MO concentrations Figure 4. Time series of dust concentrations and of cultivable fungi and bacteria during three major dust events. (a) March May 1996; (b) January March 1997; (c) May June Units: fungi and bacteria, CFU m )3 ; dust, lg m )3.

8 8 increased sharply during the dusty period and dropped to background levels once it ended on 28 February; they remained low until dust concentrations jumped again on 4 March. During this event bacteria levels were substantially higher than during the spring 1996 example (Figure 4a). Here too we note that although MOs are associated with dust episodes, they are not tightly correlated to dust concentrations. For example, peak, dust concentrations were measured on March when MOs were at intermediate levels. In contrast bacteria peaked on 9 March, just before the peak dust concentrations. The highest concentrations of dust, bacteria and fungi were measured in early June 1997 (Figure 4c). During much of this dust event bacteria and fungi concentrations were roughly equal and the peaks-and-valleys of the three species were closely matched. The results in 1997 are of particular interest because of the El Nin o that began in that year, one of the strongest on record in the 20th century (McPhaden, 1999). Dust concentrations in 1997 were among the very highest measured over the entire record at Barbados, starting in 1965 (Prospero, 1996; Prospero and Lamb, 2003). Prior to 1997, comparably high dust concentrations were only measured in when a very strong El Nin o was in effect (Prospero and Lamb, 2003) Transport characteristics of African dust, bacteria and fungi The close match (in a broad sense) between the temporal variability of dust with that of fungi and bacteria suggests that they are transported from North Africa. Dust emerges from North Africa in pulses. During the summer, dust is generally carried behind easterly waves. The highest concentrations are usually found in a layer that extends from the top of the marine boundary layer (MBL), at about 1 km over the western Atlantic, to an altitude of 3 4 km (Karyampudi et al., 1999; Reid et al., 2002, 2003). The elevated layer, because of its desert origin, is hot and dry, making it easy to identify in meteorological soundings. Because of these properties it is commonly referred to as the Saharan Air Layer (SAL) (Carlson and Prospero, 1972; Karyampudi et al., 1999). It is notable that the seasonal period of dust transport coincides with the hurricane season in the tropical Atlantic; indeed, the SAL and the entrained dust may play a role in modulating the frequency and intensity of tropical cyclones in the region (Dunion and Velden 2004). The African dust plume undergoes a seasonal displacement that follows the seasonal changes in large-scale circulation over the Atlantic. During the winter months the dust plume is carried in the low latitudes to the NE coast of South America (Prospero et al., 1981; Swap et al., 1992). In the summer months it reaches its northernmost position; then the plume passes Barbados and extends deep into the Caribbean, the Gulf of Mexico, and the southern and eastern US. The seasonal oscillation of the dust plume is consistent with the seasonal cycle of dust concentrations as measured in Barbados (Prospero and Lamb, 2003), Miami (Prospero, 1999), and Bermuda (Arimoto et al., 1992, 1995). While dust transport is greatest in the summer at Barbados, dust is occasionally transported during winter and more frequently during spring (Prospero and Lamb, 2003) as seen in Figures 3 and 4. Various satellite aerosol products (Herman et al., 1997; Husar et al., 1997; Kaufman et al., 2002) enable us to trace individual dust outbreaks from the time the dust clouds emerge from the coast of West Africa until they reach the western Atlantic and Caribbean about one week later. A number of these are available on the web in near-real time. The Total Ozone Mapping Satellite (TOMS) absorbing aerosol product (Herman et al., 1997) is particularly useful because it provides a measure of the atmospheric loading of UV-absorbing aerosols (i.e., mineral dust and soot from anthropogenic and natural combustion sources) on a global scale on a daily basis. In contrast to most aerosol-sensing satellites which are most effective over water surfaces, TOMS provides observations over both water and land surfaces and thereby yields information on sources (Prospero et al., 2002). An example of the TOMS product is given in Figure 5 for 13 and 14 June These show a plume of dust extending from the west coast of Africa to the Caribbean. On these 2 days the concentration of dust and MOs increased sharply on Barbados (Figure 4c). During the summer months on most days TOMS shows large areas of the tropical Atlantic covered with dust.

9 Nonetheless satellite observations in general, and TOMS in particular, when coupled with meteorological data and air mass trajectories (as discussed in the following section) provide strong support for the interpretation of our data in terms of source and transport paths. Unfortunately, the NIMBUS 7 detector has degraded over the past several years and current performance is marginal. However a new replacement satellite, Aura-OMI, was launched in July Dust variability and transport trajectories 9 Figure 5. TOMS absorbing aerosol distributions June 1997; Top: 13 June; Bottom: 14 June. Color code: Red high concentrations of absorbing aerosols (i.e., dust and smoke); Yellow intermediate; Gray low. aerosols/aerosols.html. The TOMS images in Figure 5 also provide an indication of the dominant dust sources and, possibly, MOs. On 13 and 14 June 1997 (Figure 5), the red-coded (maximum aerosol index) areas over West Africa suggest major dust activity in Mauritania and northern Mali, regions known to contain major dust sources (Prospero et al., 2002). While satellite images such as those from TOMS are useful in interpreting our data records, they do have limitations. TOMS does not produce reliable data in the presence of cloud and cloud becomes increasingly more common over the western Atlantic and the Caribbean. Moreover, TOMS requires relatively high concentrations of dust to yield an unambiguous response (Chiapello et al., 1999). Indeed, on most occasions the concentration of dust over the western Atlantic and Caribbean is relatively low compared to the situation depicted in Figure 5 and the TOMS product does not show the presence of aerosols as shown in the examples. Indeed, at low dust concentrations, the TOMS response is attributable largely to the presence of other types of absorbing aerosols, e.g., black carbon (Chiapello et al., 1999). Air parcel back trajectories provide insights on the source of aerosols arriving at Barbados. During the summer months, when dust is present almost every day, trajectories from within the climatological SAL altitudes generally lead back to the coast of Africa 5 7 days earlier. Figure 6 (NOAA ARL HYSPLIT4 1997) shows back trajectories on 4 June 1997 when the concentrations of dust, fungi and bacteria reached a simultaneous peak. Trajectories are shown for three starting altitudes over Barbados: 500 m, within the MBL at Barbados; 2000 m, within the SAL which typically extends from about 1000 to about 3000 m over Barbados; 4000 m, normally above the top of the SAL. During the 10-day back-trajectory period, the 500 m trajectory never touches land; it sinks from the middle troposphere over the northwestern and central North Atlantic. The 2000 m trajectory leads back to Africa crossing the coast in Senegal 7 days earlier; it then hooks north into Mauritania where some of the world s most intense dust sources are located (Prospero et al., 2002). The 4000 m trajectory crosses the coast of Africa 8 days earlier passing over the Gulf of Guinea coastal states, a region where there are no major dust sources. The example in Figure 6a for 4 June 1997 is typical of the summer months. During the 30 days of June 1997, only three 500 m trajectories traced back into Africa; the mean back-transit time was 9.0 days. Only one other 10-day 500-m trajectory (8 June) crossed onto land, passing over the NE coast of Arctic Canada 10 days earlier. Of the m trajectories, 19 traced back to Africa with a mean transit time of 6.6 days while 15 of the 4000 m trajectories crossed into Africa with a mean transit time of 6.9 days.

10 10 Figure 6. HYSPLIT back trajectories from Barbados, West Indies, at three altitudes: 500, 2000, 4000 m. (a) 4 June 1997; (b) 15 June Trajectories obtained from the NOAA ARL READY web site: In each panel symbols mark 24-hour time periods; the lower portion of each panel shows the altitude history of the back trajectories starting over Barbados at 500 m, 2000 m and 4000 m. The fact that so few 500 m trajectories reach Africa would seem to be inconsistent with the fact that we measured substantial concentrations of dust in the MBL at Barbados almost every day in June The dust that we measure at the surface in Barbados is not necessarily transported across the Atlantic at this low altitude. As stated earlier, the primary transport path of dust is in the SAL (Reid et al., 2002, 2003). Aerosols initially present in the SAL could settle into the MBL or be transferred there through convective mixing associated with small trade-wind clouds. This transport mode would not be reflected in backtrajectory calculations. In fact the monthly mean dust concentrations measured at the surface on Barbados are highly correlated with the monthly mean column aerosol optical depth (r ¼ 0.93) but the day-to-day concentrations are not (Smirnov et al., 2000). It should be noted, however, that the computation of back trajectories in the tropical Atlantic is difficult because of the dearth of data over the ocean and the complex meteorology in the tropics. They become less dependable the longer the backcalculation times. Thus trajectories should not be interpreted too closely. As an example, in Figure 6b we show the back trajectories for 15 June 1997 when dust and MOs were peaking. The looping paths reflect the influence of a depression passing to the north of Barbados, complicating the calculation of trajectories. Indeed on the following day, 16 June, all three paths looped in a similar fashion with none tracing back to Africa; yet dust and MO concentrations remained moderately high. In contrast back trajectories computed from a point several degrees to the east of Barbados do track back to Africa. Such discrepancies emphasize the point made earlier that the dearth of meteorological data in the tropical Atlantic and the complex meteorology make such calculations difficult. Despite these shortcomings, the trajectories do present a generally consistent picture of aerosol transport paths over the region and their seasonal variability. In contrast to the summer trajectory picture, trajectories during the winter and through much of the spring generally trace back to the North Atlantic, many hooking westward to North America or sometimes eastward to Europe. These winds sometimes carry substantial concentrations of pollutant SO 4 = and NO 3 ) (Savoie et al., 1992, 2002) but, as we show here, they carry no organisms cultivable on our media. A good example of the effect of changes in air mass trajectory is found in March through April 1996 (Figure 4a). Until about mid-march, dust and fungi concentrations are at minimum values and nss-so 4 = is close to background levels for this

11 11 region. On 22 March, nss-so 4 = increases sharply and rises to a plateau at about lg m )3 which extends to 12 April and then drops sharply. In contrast, mineral dust and fungi concentrations remain at minimum values until 31 March when they rise sharply and remain high into early April. Back trajectories show that on 30 March, immediately prior to the arrival the dust pulse, air parcels at 500 m over Barbados trace back to northern North America a week earlier; the 2000 and 4000 m trajectories hook back toward the Gulf of Mexico. The altitude tracks show that all trajectories descended from moderately high altitudes, over 4000 m. The concentrations of nss- SO 4 = (Figure 3) and NO 3 ) (not shown) are relatively high along with sea-salt but dust and fungi remain extremely low. Dust peaks on 3 April (Figure 3a). On that day the 500 m (MBL) back trajectory traces back to the African coast a week earlier and penetrates into southern Mauritania and Mali, a region known to have major dust sources (Prospero et al., 2002). The 2000 m trajectory passes somewhat further to the south while the 4000 m trajectory crosses into the South Atlantic, never touching land in its 10-day history. Nss-SO 4 = and NO 3 ) remain high during this period reflecting the presence of soil materials and also pollutants, most likely from European sources (Savoie et al., 1992). On 14 April, when dust levels are again low, the trajectories once again trace back to the North Atlantic and North America. A similar trajectory scenario holds for the February March dust event (Figure 4b). The task of calculating back-trajectories is especially difficult in the spring months when Barbados lies on the northern boundary of the trans- Atlantic dust transport belt (Husar et al., 1997). At such times it is not unusual to measure high concentrations of dust and MOs on days when the trajectories do not trace back to Africa but, rather, wander aimlessly in the tropical subtropical Atlantic. Ultimately the best confirmation of an African origin is the presence of dust in filter samples; the dust produces a distinct red-brown color that is quite visible to the naked eye Comparisons with other regions It is difficult to compare literature reports of airborne bacteria and fungi concentrations in different regions because of the wide variety of collection and culturing practices used (Comtois and Isard, 1999). Nonetheless, a summary of measurements (Lighthart and Stetzenbach 1994) shows concentrations in rural (non-agricultural) regions ranging from 1 to 3802 CFU m )3 ; typical ranges are from low units to several hundred CFU m )3. Lacey and Venette (1995) report typical peak concentrations of about m )3. With regard to bacteria, Lacey and Venette (1995) find that concentrations are higher in cities (up to 4000 CFU m )3, average 850) than in rural areas (up to 3400 CFU m )3, average 99). Muilenberg (1995) cites concentrations of cultivable bacteria outdoors in the range of CFU m )3. Thus bacteria and fungi concentrations in Barbados trade winds are at the extreme low end of the range of continental values and times lower than typical continental values Genera and species characteristics Table 1 shows the frequency of occurrence of colony-forming bacteria and fungi in based on identifications of a total of 13,410 colonies. The colony counts were evenly split between fungi (48.8%) and bacteria (51.2%), the latter almost exclusively Bacillus. The dominant fungus was Mycelia sterila which comprised 48% of the fungal colonies. Arthrinium (22%) and Periconium (12%) were also common along with unidentified black/gray (7%) and brown/tan (3%) colonies. Review articles attempt to compare species distributions in different environments. Lacey (1991) states that Cladosporium is the most abundant spore type on an annual basis in temperate and most tropical regions although other spores might be regionally dominant in some seasons; Alternaria spores are the second most abundant overall and in warm, dry regions can exceed the concentrations of Cladosporium. In tropical regions Curvularia and Nigrospora sometimes make large contributions and Aspergillus species are particularly characteristic of humid tropical regions. Lighthart and Stetzenbach (1994) find that the most prevalent genus of fungi at remote sites is often Cladosporium although Alternaria is frequently observed. Alternaria and Cladosporium are dematiaceous Hyphomycetes (asexual filamentous fungi); both are pigmented. Moreover Cladosporium spores are often clustered.

12 12 There is little information on airborne MOs in the region of Africa from which Barbados dust originates. Also the limited literature from arid regions is based on a variety of techniques that makes direct comparisons questionable. Table 1 summarizes literature from arid regions in Africa and the Middle East. In northern Nigeria, somewhat to the south of the major dust sources that impact Barbados, cultures were dominated by Cladosporium (36.8%) and Curvularia (25.1%) (Dransfield, 1966). In Kuwait, the most common fungal spore was Cladosporium (66%) followed by Ustilago (8%), Alternaria (4%), Helminthosporium (3%), Basidiospores (4%), and Mycelial fragments (6%) (Davies, 1969). Halwagy (1989) measured the seasonal concentration of spores at three sites in Kuwait in He found that Cladosporium was most common by far (ca. 60%). Next was Ustilago, 9 10% and Alternaria, about 6%. Al-Subai (2002) made daily measurements of airborne fungi at Doha, Qatar, over a 1-year period. He found that Cladosporium was most common (40.1% of total fungi) followed by Alternaria (21%) and Ulocladium (9.2%). Two fungi that are usually dominant in soils, Aspergillus and Penicillium, were relatively minor in air (4.3% and 3.95%, respectively). In a 1-year study at two sites in Riyadh, Saudi Arabia, (Al-Suwaine et al., 1999), Cladosporium was dominant at about 40% of the total airborne fungi. Penicillium was second in frequency (23% and 14%) and Aspergillus about the same (18.7% and 17.2%). Alternaria and Ulocladium were both roughly about 5%. Al-Suwaine made measurements by pumping air onto agar plates using two nutrients, one of which, Sabouraud s was used in our work. Of the reports cited in Table 1, only this one yields quantitative volume concentrations which enable comparisons to our data. They report mean monthly concentrations in the range of about CFU m )3. Thus, to the extent that comparisons are warranted, the airborne concentrations of cultivable fungi in Africa and the Middle East and the relative concentrations of the various genus/species appear to be quite different from those observed in Barbados trade winds. The dominance of Cladosporium is apparent in these studies regardless of location whereas in Barbados trade winds it is a minor component. It is notable that Aspergillus makes up only a very small fraction of the fungal spores. Aspergillus is widely distributed and is common in soil and on decaying vegetation, dust, and other organic debris (Levetin 1995). Many species of Aspergillus are tolerant of temperatures at or above 37 C (Levetin 1995) which might lead us to expect relatively high abundances in our samples. Nonetheless, in Barbados trade winds concentrations were low, only 1.1%, spread across 5 species (Table 1) Comparison with airborne fungi over inland Barbados We made measurements at inland sites on Barbados to compare gross concentrations and genera/species profiles with those obtained at our coastal site. Measurements were made concurrently over a period of 2 weeks at the tower on the east coast, where all the samples discussed thus far in this paper were taken, and at the University of the West Indies (UWI) Cave Hill campus, 22 km to the west of the tower (Figure 1). A few samples were also taken at a site approximately 1 km inland from the east coast and 5 km south of our tower site (Figure 1). Results are shown in Table 2. The inland samples were only 15 min duration because of the high concentrations of spores and bacteria. Thus the inland concentrations measured on any one day are not necessarily directly comparable to those at the tower where samples are nominally collected over an entire day. At the tower site the mean concentrations of fungi and bacteria were 0.36 and 0.13 CFU m )3, respectively. At UWI the values were 213 and 120 CFU m )3, 591 and 923 times the tower values, respectively. Although only two filter samples were taken at the site 1 km inland, even here the concentrations are well over 100 times higher than the tower values. Samples were also collected with a Rotorod sampler, a whirlingarm impactor, and the spores counted by microscope (Frenz et al., 1995). The spore concentrations at the UWI site were in general about 100 times greater than at the tower site (Table 2). Our inland sampling results are consistent with the expectation that MO concentrations over the ocean are low and that MOs from local vegetation and soil sources will overwhelm advected MOs even at short distances inland (Lighthart 2000). The fungal species profile in the inland samples

13 13 Table 2. Concentrations of bacteria, fungi, and spores collected at three sites on Barbados AEROCE tower (on east coast) University of the West Indies, Cave Hill (22 km from East Coast) House (1 km from East Coast) Conc. (CFU/m 3 ) Rotorod samples Conc. (CFU/m 3 ) Rotorod samples Conc. (CFU/m 3 ) Rotorod samples Date On Fungi Bacteria Total count Conc./m 3 Fungi Bacteria Total count Conc./m 3 Fungi Bacteria Total count Conc./m 3 7 July July July July July July July July July July July July July July July July July July Average

14 14 was also quite different from the mean profiles observed at the tower and from the 2-year tower means (Table 1). The dominant fungi were Cladosporium (34%), Aspergillus (A. niger, 14%; A. flavus, 3%), Bipolaris (14%), Curvularia (13%), and Penicillium (8%). We caution that a direct comparison with the tower samples in this short experiment is not necessarily valid because of the low colony counts in the tower samples. Also a close comparison with the 2-year means is probably not warranted because of the large seasonal variability in fungal species concentrations that one might expect over the island. Nonetheless the inland results are consistent with the generalizations cited above that Cladosporium is the most abundant spore type on an annual basis in temperate and most tropical regions (Lacey 1991; Lighthart and Stetzenbach 1994) and that Curvularia and Aspergillus are characteristic of humid tropical regions (Lacey 1991). 4. Discussion 4.1. Survival mechanisms The MOs that we measure in aerosols at Barbados are ubiquitous in soils and plants around the world (Lighthart and Stetzenbach, 1994; Muilenberg, 1995). This raises the question: Why are MOs associated with African dust viable after a week or more in the atmosphere, while those from other regions apparently are not? Also, why are the species/genera profiles in our samples so different from those observed in the air over the continents? There are many environmental factors that can stress and kill MOs (Cox, 1989; Marthi, 1994; Muilenberg, 1995); among the more important are desiccation, heat, and UV radiation (Aylor, 1999). There are very few data on the survival of spores exposed to sunlight and even fewer studies of the effect of exposure to UV (Aylor, 1999). The few data that are available show a wide range of exposure sensitivities (Aylor, 1999) and suggests that UV exposure during transport would have a great impact on MO survival probabilities. Because the major dust source regions are arid (Prospero et al., 2002) we would expect these spores to be relatively resistant to these various stresses. Aerobic gram positive Bacillus species form highly resistant endospores. Many fungal species that we identify in Table 1 produce darkcolored spores which would make them more resistant to solar radiation (Tong and Lighthart, 1998). Al-Subai (2002) in a study of airborne fungi in Qatar noted that the predominant species (Cladosporium, Alternaria, and Ulocladium) are dark-colored. Al-Subai also cites literature that reports that dark-colored conidial fungi are prevalent in sandy soils of the Sahara, Egypt and the Sonoran deserts. Another factor might be that the thick clouds of dust attenuate the UV flux which might otherwise kill the organisms (Liu et al., 1991; Mims et al., 1997). At Barbados during spring and summer, dust accounts for 60% of the light attenuation in the mean and much more during dust events (Li et al., 1996). The monthly mean column aerosol optical depth ranges between 0.2 and 0.3 at 440 nm and is highly correlated (r ¼ 0.93) with monthly mean dust concentrations measured at the surface (Smirnov, 2000). Dust concentrations and dust optical depths (and UV attenuation) are much higher over the eastern Atlantic and over West Africa (Hsu et al., 1999) and thus would afford much more shielding against UV. Another effective UV protective mechanism would be the shielding of individual spores by dust particles. The spores, many of which are a few micrometers diameter, might be covered with fine soil particles or they might be attached in a niche in large particles or clumps. On Barbados, under typical dusty conditions, about 10% of the particle mass is in the size fraction greater than 10 lm aerodynamic diameter (Li-Jones and Prospero, 1998); on some occasions several percent of the mass exceeds 20 lm (Prospero et al., 1970; Li-Jones and Prospero, 1998). Thus during a typical dust event, a cubic meter of air contains several hundred to a thousand particles above 10 lm and ten s of particles above 20 lm. These could easily accommodate spore particles during transit to Barbados. Although our study shows that cultivable MOs are only found in the presence of mineral dust, we can not say whether the organisms were directly associated with individual dust particles or whether they were independently suspended in the air. If the organisms were associated with individual dust particles, it would suggest that the organisms may have been derived from soils directly when the dust was mobilized by winds. As previously stated, many of the fungi that we observe on Barbados are

15 15 found in soils. However there is very little information on the soils in the dust source regions in North Africa. Also, differences in sampling and culturing techniques make comparisons difficult. Nonetheless Dransfield (1966) in his study in northern Nigeria cultured air samples and compared the results with cultured extracts from local soils. The relative genera concentrations in air were quite different from those in soils. The dominant genera in soils were Penicillium and Aspergillus, which each made up 33% of all colonies; in contrast these were minor in air in Nigeria (1.8% and 1.7%, respectively). In Qatar, Alternaria and Cladosporium, were the most common genera in air (40.1% and 21%, of the total) whereas they accounted for only 4.06% and 2.8% of the total soil fungi (Al-Subai 2002). Our work does not preclude the possibility that the MOs could have been derived from sources completely different from the dust source regions. The MOs could have been advected over the dust source region and subsequently became mixed with deflating soil dust. Alternatively, MOs could have been injected into dusty air masses as they passed over West Africa on their way to the Atlantic. These two scenarios would be consistent with our observation that bacteria and fungi concentrations are essentially uncorrelated with dust and with the fact that some major dust peaks are not associated with any substantial increase in MO concentrations (e.g., as shown in Figures 3, 4). If the MOs are derived from non-soil sources, it is possible the organisms could subsequently become attached to soil particles during transit, most likely when air parcels are processed through clouds. It is clear, however, that even in relatively arid regions plants are major sources of airborne fungi. In the African and Middle-East work cited here, it is usually noted that periods of high spore concentrations were linked primarily to the rain vegetation cycle. Another possible explanation for the viability of MOs with African dust events is related to the transport path of the MOs. Although the major transport path for African dust is in the Saharan air layer that lies above the marine boundary layer, a substantial fraction of the transport can take place in the MBL where the relative humidity is high, typically above 70% (Reid et al., 2002), minimizing desiccation effects in MOs. In contrast, air trajectories that arrive in Barbados from the North Atlantic (and North America and Europe) often sink from the middle troposphere where temperatures and relative humidity can be very low and UV fluxes very high, factors that could kill MOs or render spores nonviable Comparison with other studies of LRT As previously stated, there have been very few studies of microbe LRT and none which would allow the association of MOs with a source region. Recently Griffin et al. (2001), in a study on St. John in the Virgin Islands in July 2000 using sampling and culturing techniques similar to ours, found high concentrations of viable MOs, including pathogens, which they associated with the presence of African dust. Similar results are reported in a more recent paper, Griffin et al. (2003). They measured concentrations much higher than those obtained by us on Barbados. Their mean bacteria concentrations were 308 times greater than our July means and the fungi concentrations 89 times greater. Moreover, they identify a wide variety of MOs that are very different from those found by us at our tower. Finally, we note that Griffin et al. never made any physical measurements of dust concentrations; they simply inferred the presence (or absence) of dust based on visibility and on the TOMS aerosol product which, as we noted above, is highly ambiguous over the Western Atlantic except at very high dust concentrations (Chiapello et al., 1999). We note that their sampling site was located at Lind Point on the extreme western end of the island of St. John. Thus under typical summer trade wind conditions, winds must traverse the 15-km length of the island as well as a number of small islands lying to the east. Our experiences in sampling at inland sites on Barbados suggest that their samples are highly impacted by local sources. In fact their results at Lind Point are quite comparable to those obtained by us at our inland sampling site at UWI (Table 2) both from the standpoint of the concentrations and the species/genera profiles Implications of the LRT of MOs, climate change, and health Although we only present a 2-year record here, we see a great difference in the concentration of bacteria and fungi in those 2 years. The large