Global traffic and disease vector dispersal

Andrew J. Tatem*†‡, Simon I. Hay*†, and David J. Rogers*

*Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, United Kingdom; and † Public Health and Epidemiology Group, Centre for Geographic Medicine, Kenya Medical Research Institute, P.O. Box 43640, 00100 Nairobi GPO, Kenya

Edited by Kenneth W. Wachter, University of California, Berkeley, CA, and approved February 21, 2006 (received for review September 27, 2005) The expansion of global air travel and seaborne trade overcomes exotic mosquitoes on many Pacific islands, including Hawaii (12, geographic barriers to insect disease vectors, enabling them to 17). Modern container ships are known to have introduced move great distances in short periods of time. Here we apply a numerous alien mosquito species, including the Asian tiger coupled human–environment framework to describe the historical mosquito, Aedes albopictus (18), to new areas (12). Ae. albopictus spread of Aedes albopictus, a competent mosquito vector of 22 is of medical and public health concern because it has been arboviruses in the laboratory. We contrast this dispersal with the shown in the laboratory to be a highly efficient vector of 22 relatively unchanged distribution of gambiae and ex- arboviruses, including dengue, yellow, and West Nile fever amine possible future movements of this malaria vector. We use a viruses (19, 20). In the wild, however, its efficiency as a vector comprehensive database of international ship and aircraft traffic appears to be generally low (19, 20), although it has been movements, combined with climatic information, to remap the implicated in recent dengue fever outbreaks in the absence of the global transportation network in terms of disease vector suitability principal vector, Ae. aegypti (20, 21). From its Old World and accessibility. The expansion of the range of Ae. albopictus distribution reported in 1930 (Fig. 1), Ae. albopictus expanded its proved to be surprisingly predictable using this combination of range first to the Pacific Islands (20) and then, within the last 20 climate and traffic data. Traffic volumes were more than twice as years, to other countries in both the Old World and the New high on shipping routes running from the historical distribution of World (Fig. 1) (20, 22), principally through ship-borne trans- Ae. albopictus to ports where it has established in comparison with portation of eggs and larvae in tires (23). routes to climatically similar ports where it has yet to invade. In Although distance may no longer represent a significant contrast, An. gambiae has rarely spread from Africa, which we barrier to vector movement, climate at the point of entry still suggest is partly due to the low volume of sea traffic from the presents a fundamental constraint to establishment, because continent and, until very recently, a European destination for most small invertebrates are very sensitive to the weather. Here we flights. combine data on global air and sea traffic volumes with climatic data to examine the effects of these constraints on global disease Aedes albopictus ͉ air travel ͉ ͉ biological invasion ͉ vector dispersal and establishment. By remapping the global shipping transport network in terms of climatic similarity, we examine both how predictable was the spread of Ae. albopictus and what hroughout history the opening of travel and trade routes might be the potential for future dispersal of both Ae. albopictus Tbetween countries has been accompanied by the spread of and An. gambiae. microbes and their vectors (1). Significant events include the 14th- to 17th-century pandemics, the 19th-century chol- Results era epidemics, and the escape of Aedes aegypti from West Africa, Fig. 2 displays climatic dendrograms for the global shipping facilitating epidemics in North American port cities network (Fig. 2A) and the global air travel network (Fig. 2B). in the 19th and early 20th centuries (2). The most devastating The dendrograms are climate-based phenetic trees that repre- disease vector introduction in recent history was that of Anoph- sent the global transport network remapped in terms of suit- eles gambiae (3, 4), the most important malaria vector in Africa, ability for each disease vector. The cutoff level on the seaports to northeastern Brazil in 1930. Although malaria was already dendrogram that included 90% of ports within the Old World endemic in the area, the greater vectorial capacity of this distribution of Ae. albopictus encompassed a single branch. This mosquito species, relative to local species, sparked branch consisted of two major subbranches reflecting the trop- falciparum malaria epidemics costing 16,000 lives (5) before ical and temperate Ae. albopictus races (24), with the remaining eradication of An. gambiae by an effective, yet costly, control 10% of ports displaying quite distinct environments (Mormugao, program (6). New-Mangalore, and Kuching). Ninety percent of airports During the last 50 years air travel passenger numbers have within the preexpansion distribution of Ae. albopictus were grown by nearly 9% per annum (7); shipping traffic has increased encompassed in a single branch of the airport dendrogram, but by Ͼ27% since 1993 (8). This growth is associated with rising temperate and tropical races were again distinguishable within numbers of cases of biological invasion. Gradual mixing of the this branch. world’s biota represents a global problem, with consequences From Ͼ25,000 possible sea routes and Ͼ6,000 air routes, Table ranging from native species’ decline to threats to human health 1 shows the top 10 shipping and air travel routes identified as (9). Biological invasion is a multistep process involving initial having the highest relative invasion risk (defined in Materials and dispersal, establishment, and spread (10, 11). Invasion research Methods) for Ae. albopictus and whether or not this species has has increased considerably over the past 15 years, yet the initial been recorded at the arrival port. We attempted to quantify the dispersal phase, on which the other stages rely, remains the least relative importance of sea traffic volume and local climate in the studied (9). Recent biological invasions have involved both establishment of Ae. albopictus by examining all 68 ports within disease vectors (12) and the expansion or reemergence of the invasion risk dendrogram branches. Some (n ϭ 21) ports vector-borne diseases (13). Modern transport facilitates the movement of disease vectors farther and faster than ever before. Aircraft have been proven Conflict of interest statement: No conflicts declared. capable of transporting disease vectors through experiments This paper was submitted directly (Track II) to the PNAS office. (14), the phenomenon of airport malaria (15), and incoming Freely available online through the PNAS open access option. flight surveys (16). Air travel is also suspected to be directly ‡To whom correspondence should be addressed. E-mail: [email protected]. responsible for the dispersal and consequent establishment of © 2006 by The National Academy of Sciences of the USA

6242–6247 ͉ PNAS ͉ April 18, 2006 ͉ vol. 103 ͉ no. 16 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0508391103 Downloaded by guest on September 27, 2021 Fig. 1. Old World distribution of Ae. albopictus (red), countries reporting established breeding populations of Ae. albopictus in the last 30 years (orange), and countries reporting Ae. albopictus interception at ports (yellow).

were within the original range of Ae. albopictus, but, of the other analysis suggests that the combination of traffic volumes and 47, to date just over half are in regions colonized or where climatic suitability can explain a great deal of the global spread breeding populations have been found. Within this group of 47 of Ae. albopictus and that, when the climatic suitabilities of sites ports, the average climatic distances of the invaded (n ϭ 24) and are similar, shipping volume alone appears to determine invasion noninvaded (n ϭ 23) ports were identical, but average sea traffic probability. volumes were significantly (2.43 times) greater in the former Although air travel is not implicated in the spread of Ae. than the latter (41.84 ship visits per annum per route for invaded albopictus, 6 of the 10 highest-risk air transport routes (Table 1) ports and 17.24 for noninvaded ones; t ϭ 2.343, P ϭ 0.0024). link the original range of Ae. albopictus with Honolulu in Hawaii, Fig. 3 shows the 10 sea traffic routes identified with the where this species has become established (33), whereas the strongest risk factors for importation and establishment of each remaining destinations have all reported interceptions of this of four principal members of the An. gambiae complex. The mosquito. Recent work suggests that air travel was indeed the results of the air traffic analysis for the African mosquitoes are most likely route of invasion for Ae. albopictus to Hawaii (17). not displayed here because the current European bias in flight Details on Hawaiian ports were not available in the sea traffic routes from Africa and the climatic differences between the two database. continents mean that few risk routes were identified. Although the efficiency of Ae. albopictus as a disease vector Discussion remains unclear, the An. gambiae complex of African mosquitoes are proven efficient vectors of P. falciparum malaria. Their There is excellent correspondence between the predicted top 10 risk establishment outside Africa, which has been the subject of routes in Table 1 and the global spread or interception of Ae. recent speculation (34), could have devastating consequences albopictus. Three of the top 10 shipping routes run from Japan to (6). The results in Fig. 2 show that An. gambiae sensu stricto and the southeast United States, where some of the earliest breeding Anopheles melas, from the hotter, wetter parts of tropical West populations of Ae. albopictus were found and identified as origi- and Central Africa, are more likely to thrive in climatically nating from Japan (12). Genoa, the destination of two more routes ͞ from Japan, was one of the earliest European cities to report Ae. equivalent zones in Southeast Asia and Central South America.

albopictus, which was established there by 1990 (25); the vector has In contrast, Anopheles arabiensis prefers drier climates, which ECOLOGY since become the most important local nuisance mosquito (20). Of match those of the Middle East and Canary Islands, whereas the remaining routes in Table 1, interceptions of Ae. albopictus in Anopheles quadriannulatus prefers relatively cooler habitats in tire shipments from Japan to Australasia have been made in both southern Africa that are matched most closely by those else- Brisbane (26) and Auckland (27), but the species appears not to where in South Africa, Mediterranean Europe, and Japan. In have established, probably an endorsement of the strict local each case, the combination of local climate and volume of sea inspection and fumigation policies (12). There is no documented traffic determines the areas at greatest risk of invasion. Unlike evidence of invasion of Fraser, Canada, by Ae. albopictus. Given the the Ae. albopictus risk routes, however, there is no evidence of climatic similarity of ports in Japan and Fraser and the level of An. gambiae invasion at any of these destinations to date. This shipping traffic between the two, Fraser should be considered at absence of invasion may be partly explained by the relatively low high risk, especially because Ae. albopictus has already been inter- sea traffic volumes originating from most African ports, which cepted at nearby Seattle (28). are only 1͞10th (An. melas and An. quadriannulatus inhabited Traffic volumes within the global transport network have been areas), 1͞20th (An. arabiensis), and 1͞30th (An. gambiae sensu proposed as important in determining biological invasion suc- stricto) of the average volumes of traffic on the Ae. albopictus cess (29–31). The role of shipping depends on a range of factors, routes in Table 1. These figures appear to reinforce the conclu- such as ship type and destination port (32). Although it has been sion about the importance of traffic volume, although An. suggested that invasion pressure (the number of arriving prop- gambiae sensu lato may also be less prone to move than Ae. agules) is not a simple function of total ship arrivals, the present albopictus because it is less well adapted to anthropogenic

Tatem et al. PNAS ͉ April 18, 2006 ͉ vol. 103 ͉ no. 16 ͉ 6243 Downloaded by guest on September 27, 2021 Fig. 2. Climatic similarity dendrogram for the major seaports of the world (A Inset) and the major international airports of the world (B Inset), with a sample of the branches relating to the spread of Ae. allopictus in close-up on each.

breeding sites, such as tires (3, 4), and is less frequently found in siderably increasing the risk of introducing malaria vectors from urban areas (35). Africa. In contrast to the sea routes, which originate in Africa and There are currently no confirmed examples of vector estab- have distant destinations in ecologically equivalent environ- lishment arising from air travel (12), although there is circum- ments, the air travel network from the African continent is stantial evidence that this has occurred (12, 16, 36). This mode primarily to European destinations, outside of the climatic of transport, however, may be more important in delivering the suitability limits of An. gambiae s.l. The air transport destinations disease, rather than the vector, via infected human passengers most at risk are, therefore, generally to be found within the (37, 38). Furthermore, the continuing rise in imported and continent itself or are very close to it. Thus, we identify existing airport malaria cases (39, 40), the spread of insecticide resistance potential risks of air transfer of An. arabiensis to the Middle East in mosquitoes, and antimalarial drug-resistant malaria parasites and of An. quadriannulatus from Mozambique to Southern (41, 42) are all reflections of an expanding global air transport Europe (Lisbon). The air transport climate dendrogram also network, increased travel to malarious countries (43), and identifies climatic similarities to Africa in both South America declining aircraft disinsection (16). and Southeast Asia, but these were not linked by any direct Successful biological invaders can be difficult to predict flights in the database used for this analysis. Recently, however, given the lack of evidence of a universal trait related to long-haul flights from Africa direct to Washington, DC; Beijing; invasion (44). The identification here, however, of the princi- Hong Kong; and Bangkok, Thailand, have been opened, con- pal routes of movement of Ae. albopictus from thousands of

6244 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0508391103 Tatem et al. Downloaded by guest on September 27, 2021 Table 1. Ae. albopictus shipping and air travel risk routes Ae. albopictus Ae. albopictus Rank From To found? established?

Shipping 1 Chiba, Japan New Orleans Y Y 2 Chiba, Japan Genoa, Italy Y Y 3 Chiba, Japan Fraser, Canada N ? 4 Chiba, Japan Brisbane, Australia Y ? 5 Chiba, Japan Auckland Y ? 6 Chiba, Japan South Louisiana Y Y 7 Yokohama, Japan Fraser, Canada N ? 8 Kobe, Japan Fraser, Canada N ? 9 Chiba, Japan Miami Y Y 10 Yokohama, Japan Genoa, Italy Y Y Air 1 Tokyo Narita, Japan Honolulu Y Y 2 Osaka Kansai, Japan Honolulu Y Y 3 Nagoya, Japan Honolulu Y Y 4 Tokyo Narita, Japan Seattle Y ? 5 Tokyo Narita, Japan Brisbane, Australia Y ? 6 Fukuoka, Japan Honolulu Y Y 7 Seoul, South Korea Honolulu Y Y 8 Tokyo Hareda, Japan Honolulu Y Y 9 Taipei Chang, Taiwan Seattle Y ? 10 Tokyo Narita, Japan Portland, OR Y ?

Air results are exclusive of short-haul routes to Hong Kong; Sapporo, Japan; and Pyongyang, North Korea.

possible alternatives is evidence that seaports and airports values. Each 10 ϫ 10 (Ϸ18 ϫ 18 km at the equator) grid square within the global transport network at greatest risk of future covering the seaport͞airport location was identified. To ensure spread of disease vectors can be highlighted through a com- a representative climate measure, up to eight land pixels bination of climate and traffic data. Such an approach can, surrounding each seaport͞airport grid square were identified therefore, be used to prioritize monitoring, inspection, and where possible. Any seaports͞airports located on islands too quarantine efforts. Future challenges involve incorporating small to be represented by the climate surfaces were eliminated more sophisticated climatic data and extra variables, as well as from the analysis, reducing the sample size to 241 seaports information on shipping and aircraft specifics, intraspecies and 259 airports. The selected grid square data from the competition, vector infection rates, human populations at risk, nine climate surfaces thus formed the climatic ‘‘signature’’ of seasonality, breeding site availability, disinsection, and land each port. transport routes. It is being increasingly recognized that future progress in the Hierarchical Clustering and Climatic Envelopes. The distance in fields of emerging diseases and ecology requires a holistic climatic space between each seaport͞airport signature and perspective incorporating multiple-scale social, physical, and every other seaport͞airport signature was calculated to pro- biological dimensions (45). The study of disease vector dis- duce separate seaport and airport ‘‘climatic dissimilarity’’ persal must encompass scales from the study of microbial matrices. Euclidean distance measures were used because pathogens and arthropod vectors to the global scale of eco- there were generally too few pixels for any more sophisticated logical processes affected by humans. Continuing globalization measures of environmental distance. In the few cases where will increase the relative risks of establishment of a range of they could be applied, more complex statistical measures of vectors, as will many aspects of global climate change (46). distance did not affect the results. To remap the global There is no room for complacency because, whereas quaran- ECOLOGY transportation network in terms of disease vector accessibility, tine services must succeed every time, disease vectors need each climatic dissimilarity matrix was subjected to hierarchical succeed only once to establish new bridgeheads from which to invade new regions or continents. clustering by using an agglomerative algorithm (PHYLIP V3.63, University of Washington). The clustering results were then Materials and Methods translated into dendrograms based on centroid linkage to Transport Matrices and Climatic Signatures. Estimates of the map the world’s seaport and airport connections by climatic amount of traffic among the 243 most visited international similarity. seaports in 2000 (30) yielded measures of traffic volumes on a To examine how predictable was the spread of Ae. albopictus, total of 29,403 routes. Data on total air passenger numbers each seaport and airport location was overlaid on a map of the moving between the world’s 278 principal international air- mosquito’s Old World range (Fig. 1) (20, 24, 48) and classified ports in 2000 were also obtained (OAG Worldwide). Air and as either inside or outside the historical distribution. Those sea transport matrices were produced, with each cell contain- seaports͞airports within the distribution were located on the ing the number of air passenger or ship movements between relevant dendrogram. The dendrogram branch that encom- each port. The locations of the seaports and airports were passed at least 90% of the ports was designated as defining the superimposed onto nine gridded global climate surfaces (47) limits of Ae. albopictus ’ ‘‘climatic suitability envelope.’’ In doing representing means, maximums, and minimums of tempera- this, we allowed for the fact that Ae. albopictus has both ture, rainfall, and humidity, each rescaled to the same range of temperate (diapausing) and tropical (nondiapausing) races with

Tatem et al. PNAS ͉ April 18, 2006 ͉ vol. 103 ͉ no. 16 ͉ 6245 Downloaded by guest on September 27, 2021 Fig. 3. The 10 sea traffic routes identified with the strongest risk factors for importation and establishment of each of four principal members of the An. gambiae complex. (A) An. gambiae sensu stricto (s.s.). (B) An. arabiensis.(C) An. quadriannulatus.(D) An. melas.

distinct environmental requirements and different original geo- seaports͞airports not within the historical distribution of Ae. graphical distributions (24). albopictus, but linked via a dendrogram branch within the Assuming the availability of suitable habitats and breeding environmental envelope, were therefore considered to be sites and no interspecific competition, the chance of a vector similar enough climatically for there to be a risk of vector becoming established on introduction is considered highest if establishment. the climate at any new (previously uninvaded) site is similar to Where sufficient seaport and airport data existed within the that at sites within its historical distribution (49). Those predicted range of each mosquito (50), the above procedure was

6246 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0508391103 Tatem et al. Downloaded by guest on September 27, 2021 repeated for the malaria vectors, An. gambiae sensu stricto, An. taken to represent relative invasion risk for the sites on the arabiensis, An. quadriannulatus, and An. melas. dendrograms.

Risk Matrices. The risks of invasion were quantified as a function We are indebted to Briony Boon, John Drake, Carlos Guerra, Beth of climatic similarity and traffic volume between the origin and Purse, Sarah Randolph, Paul Reiter, Dave Smith, and Robert Snow for destination seaport͞airport by rescaling independently the comments on earlier drafts of the manuscript, as well as two anonymous climatic dissimilarity and sea͞air traffic matrices to a range reviewers for valuable input. We are especially grateful to John Drake between 0 (climates of the seaports͞airports in question very for the generous supply of the sea traffic estimates. A.J.T. and S.I.H. are different; no traffic between them) and 1.0 (climates identical funded by Research Career Development Fellowship 069045 (to S.I.H.) or traffic high). The product of these rescaled measures was from the Wellcome Trust.

1. McNeill, W. H. (1976) Plagues and People (Anchor, New York). 26. Kay, B. H., Ives, W. A., Whelan, P. I., Barker-Hudson, P., Fanning, I. D. & 2. Tatem, A. J., Rogers, D. J. & Hay, S. I. (2006) Adv. Parasitol. 62, 297–347. Marks, E. N. (1990) Med. J. Aust. 153, 31–34. 3. Gillies, M. T. & De Meillon, B. (1968) The Anopheline of Africa South of the 27. Laird, M., Calder, L., Thornton, R. C., Syme, R., Holder, P. W. & Mogi, M. Sahara (South African Inst. for Medical Research, Johannesburg). (1994) J. Am. Assoc. 10, 14–23. 4. Gillies, M. T. & Coetzee, M. (1987) A Supplement to the Anopheline of Africa 28. Centers for Disease Control. (1986) Morbid. Mortal. Wkly. Rep. 35, 649–651. South of the Sahara (South African Inst. for Medical Research, Johannesburg). 29. Levine, J. M. & D’Antonio, C. M. (2003) Conserv. Biol. 17, 322–326. 5. Soper, F. L. & Wilson, D. B. (1943) Anopheles gambiae in Brazil: 1930 to 1940 30. Drake, J. M. & Lodge, D. M. (2004) Proc. R. Soc. London Ser. B 271, 575–580. (Rockefeller Foundation, New York). 31. Lockwood, J. L., Cassey, P. & Blackburn, T. (2005) Trends Ecol. Evol. 20, 6. Killeen, G. F., Fillinger, U., Kiche, I., Gouagna, L. C. & Knols, B. G. J. (2002) 223–228. Lancet Infect. Dis. 2, 618–627. 32. Verling, E., Ruiz, G. M., Smith, L. D., Galil, B., Miller, A. W. & Murphy, K. R. 7. Upham, P., Thomas, C., Gillingwater, D. & Raper, D. (2003) J. Air Transport (2005) Proc. R. Soc. London Ser. B 272, 1249–1256. Manage. 9, 145–151. 8. Zachcial, M. & Heideloff, C. (2003) ISL Shipping Statistics Yearbook 2003 (Inst. 33. Mitchell, C. J. (1995) J. Vector Ecol. 20, 44–58. of Shipping Economics and Logistics, Bremen, Germany). 34. Powell, J. R. & Coluzzi, M. (July 21, 2004) in International Herald Tribune, 9. Puth, L. M. & Post, D. M. (2005) Ecol. Lett. 8, 715–721. Opinion. 10. Elton, C. S. (1958) The Ecology of Invasions by Animals and Plants (Methuen, 35. Hay, S. I., Guerra, C. A., Tatem, A. J., Atkinson, P. M. & Snow, R. W. (2005) London). Nat. Rev. Microbiol. 3, 81–90. 11. Shigesada, N. & Kawasaki, K. (1997) Biological Invasions: Theory and Practice 36. Mangili, A. & Gendreau, M. A. (2005) Lancet 365, 989–996. (Oxford Univ. Press, Oxford). 37. Wenzel, R. P. (1996) N. Engl. J. Med. 334, 981–982. 12. Lounibos, L. P. (2002) Annu. Rev. Entomol. 47, 233–266. 38. Hufnagel, L., Brockmann, D. & Geisel, T. (2004) Proc. Natl. Acad. Sci. USA 13. Gratz, N. G. (1999) Annu. Rev. Entomol. 44, 51–75. 101, 15124–15129. 14. Russell, R. C. (1989) Travel Med. Int. 7, 26–31. 39. Muentener, P., Schlagenhauf, P. & Steffen, R. (1999) Bull. W. H. O. 77, 15. Isaa¨cson, M. (1989) Bull. W. H. O. 67, 737–743. 560–566. 16. Gratz, N. G., Steffen, R. & Cocksedge, W. (2000) Bull. W. H. O. 78, 995–1004. 40. Toovey, S. & Jamieson, A. (2003) Travel Med. Infect. Dis. 1, 167–175. 17. Kilpatrick, A. M., Gluzberg, Y., Burgett, J. & Daszak, P. (2004) Ecohealth 1, 41. Roper, C., Pearce, R., Nair, S., Sharp, B., Nosten, F. & Anderson, T. (2004) 205–209. Science 305, 1124. 4, 18. Hawley, W. A. (1988) J. Am. Mosquito Control Assoc. 1–40. 42. Okeke, I. (2004) Science 306, 2039–2040. 19. Gubler, D. J. (2003) Lancet 3, 751–752. 43. Loutan, L. (2003) Int. J. Antimicrob. Agents 21, 158–163. 20. Gratz, N. G. (2004) Med. Vet. Entomol. 18, 215–227. 44. Gewin, V. (2005) PLoS Biol. 3, 2066–2071. 21. Effler, P. V., Pang, L., Kitsutani, P., Vorndam, V., Nakata, M., Ayers, T., Elm, 45. Wilcox, B. A. & Colwell, R. R. (2005) Ecohealth 2, 244–257. J., Tom, T., Reiter, P., Rigau-Perez, J. G., et al. (2005) Emerging Infect. Dis. 11, 742–749. 46. Rogers, D. J., Randolph, S. E., Lindsay, S. & Thomas, C. (2001) in Health 22. Moore, C. G. & Mitchell, C. J. (1997) Emerging Inf. Dis. 3, 329–334. Effects of Climate Change in the U.K. (U.K. Dept. of Health, London). 23. Reiter, P. (1998) J. Am. Mosquito Control Assoc. 14, 83–94. 47. New, M., Lister, D., Hulme, M. & Makin, I. (2002) Climate Res. 21, 1–25. 24. Hawley, W. A., Reiter, P., Copeland, S., Pumpuni, C. B. & Craig, G. B. (1987) 48. Reiter, P. & Darsie, R. F. (1984) Mosquito News 44, 396–399. Science 236, 1114–1115. 49. Rogers, D. J. & Randolph, S. E. (2003) Nat. Rev. Microbiol. 1, 231–237. 25. Romi, R., Di Luca, M. & Majori, G. (1999) J. Am. Mosquito Control Assoc. 15, 50. Rogers, D. J., Randolph, S. E., Snow, R. W. & Hay, S. I. (2002) Nature 415, 425–427. 710–715. ECOLOGY

Tatem et al. PNAS ͉ April 18, 2006 ͉ vol. 103 ͉ no. 16 ͉ 6247 Downloaded by guest on September 27, 2021