spread of the tiger: global risk of invasion by the...

76

VECTOR-BORNE AND ZOONOTIC DISEASESVolume 7, Number 1, 2007© Mary Ann Liebert, Inc.DOI: 10.1089/vbz.2006.0562

Spread of the Tiger: Global Risk of Invasion by theMosquito Aedes albopictus

MARK Q. BENEDICT,1 REBECCA S. LEVINE,1 WILLIAM A. HAWLEY,1and L. PHILIP LOUNIBOS2

ABSTRACT

Aedes albopictus, commonly known as the Asian tiger mosquito, is currently the most invasive mosquito in theworld. It is of medical importance due to its aggressive daytime human-biting behavior and ability to vector manyviruses, including dengue, LaCrosse, and West Nile. Invasions into new areas of its potential range are often ini-tiated through the transportation of eggs via the international trade in used tires. We use a genetic algorithm, Ge-netic Algorithm for Rule Set Production (GARP), to determine the ecological niche of Ae. albopictus and predicta global ecological risk map for the continued spread of the species. We combine this analysis with risk due toimportation of tires from infested countries and their proximity to countries that have already been invaded todevelop a list of countries most at risk for future introductions and establishments. Methods used here have po-tential for predicting risks of future invasions of vectors or pathogens. Key Words: Insect vectors—Biodiversity—Public health—Forecasting—Risk factors—Computer simulation. Vector-Borne Zoonotic Dis. 7, 76–85.

INTRODUCTION

AEDES ALBOPICTUS, a mosquito native to Asia,has been one of the fastest spreading ani-

mal species over the past two decades. In ad-dition to the ecological problems inherent inrapid spread of any species, of particular im-portance are the serious public health risksposed by the introduction and establishment ofan aggressive pest and efficient disease vector.

Ae. albopictus has spread from its nativerange to at least 28 other countries around theglobe, largely through the international tradein used tires (Table 1 and Reiter and Sprenger1987). These recent invasions by Ae. albopictusare modern analogues to the cosmotropicalspread of Ae. aegypti on sailing vessels, aprocess documented by Tabachnick (1991). Asan inhabitant of artificial and natural contain-ers, Ae. albopictus is especially prone to inad-

vertent transport of its relatively cold-hardyand long-lived eggs, and to live in close affili-ation with humans and their dwellings andrefuse (see reviews by Hawley [1988] andEstrada-Franco [1995]). Such specific biologiccharacteristics of Ae. albopictus that are relevantto the invasiveness and ability to displace otherspecies have been reviewed elsewhere (Julianoand Lounibos 2005).

Upon establishment, Ae. albopictus may be-come merely a pest, but its potential to vectora wide range of human pathogens causes themost concern. It ranks second only to Ae. ae-gypti in importance as a vector of dengue virus,and it is a laboratory-competent vector of sevenalphaviruses (e.g., Eastern Equine encephalitisand Ross River viruses) and eight bunyaviruses(e.g., LaCrosse and Rift Valley fever viruses).Ae. albopictus is also competent to develop threeflaviviruses (Japanese encephalitis, West Nile,

1Centers for Disease Control and Prevention, Atlanta, Georgia.2University of Florida, Florida Medical Entomology Laboratory, Vero Beach, Florida.

GLOBAL RISK OF INVASION BY AEDES ALBOPICTUS 77

and yellow fever) to humans (Mitchell 1995,Shroyer 1986), and these, as well as denguevirus, transovarially to its offspring (Estrada-Franco 1995). Here, we model the global spreadof Ae. albopictus based on its native Asian rangeusing ecological niche modeling. We compareits predicted potential range to those of othersmade using environmental-limit models. Wethen compare our results to the locations ofdocumented establishments and spread overthe last 20 years. Finally, we list the countriesmost at risk for future introduction and estab-lishment of Ae. albopictus.

MATERIALS AND METHODS

Data

Mosquito occurrence. A database consisting of206 occurrence points of Ae. albopictus in Asia

was created by combining 164 records from lit-erature and 42 original points from collectionsperformed in 1989 by W. Hawley and L. Mun-stermann (personal communication). Four ofthese could not be used because they fell incoastal fringes outside of the climate layer cov-erages used. The remaining points were usedto develop the species distribution map for itsnative Asian range. These points are spatiallyaccurate to the minute and to the second inmany cases where more precise data wereavailable.

We collected additional information on ex-otic (i.e., non-Asian) arrivals or establishmentsof the species in the past 20 years by a litera-ture search and personal communication withpublic health officials. This consists of 181unique points distributed worldwide, 5574 inBrazil, and 1033 counties in the United Stateswhere Ae. albopictus has been collected andrecorded. U.S. data were converted from county

TABLE 1. COUNTRIES IN WHICH AE. ALBOPICTUS WAS ESTABLISHED BETWEEN 1979–2004

Invasion year Country Reference

1979 Albaniaa Adhami and Reiter 19981983 Trinidad Le Maitre and Chade 19831985 United States Sprenger and Wuithiranyagool 19861986 Brazil Forattini 19861988 Mexico Anonymous 19891990 Italy Sabatini et al. 19901991 Nigeria Savage et al. 1992

South Africab Cornel and Hunt 19911993 Barbados Reiter 1998

Dominican Republic Peña 19931995 Cuba Broche and Borja 1999

Guatemala Ogata and Samayoa 1996HondurasEl Salvador M. Suarez, personal communication

1997 Bolivia M. Suarez, personal communicationCayman Islands D. Malone, personal communication

Lounibos et al. 20031998 Argentina Rossi et al. 1999

Colombia Velez et al. 1998Paraguay B. Cousinho, personal communication

1999 France Schaffner and Karch 20002000 Cameroon Fontenille and Toto 20012001 Equatorial Guinea Toto et al. 20032002 Panama2003 Nicaragua Lugo et al. 2005

Greece Y-M. Linton, personal communicationIsrael L. Blaustein, personal communicationSwitzerland Flacio et al. 2004

2004 Belgium Schaffner et al. 2004Spain Nart 2004

aInfestations were detected as early as 1979.bIntroduction but not establishment was reported.

name to its centroid latitude and longitude. Alldata are available upon request from the authors.

Environmental. Eleven environmental datalayers were selected for model inclusion basedon error patterns in preliminary analyses. Tenlayers were obtained from the Intergovern-mental Panel on Climate Change (IPCC,Anonymous 1989) for the years 1961–1990: an-nual mean temperature, annual mean maxi-mum temperature, annual mean minimumtemperature, average daily temperature range,frost days per annum, topographic aspect (di-rection of pixel slope), flow accumulation(number of pixels in watershed), topographicindex (USGS-Hydro1k, an index combiningslope and flow accumulation), annual meanprecipitation, and wet days per annum. Land-use/land-cover data were from the Universityof Maryland, Department of Geography,1961–1990. Resolution (pixel size) of each envi-ronmental layer was 0.1 degrees.

GARP modeling

All models were created using the GeneticAlgorithm for Rule Set Production (GARP),which is an implementation of a genetic algo-rithm designed for predicting and modelingthe ecological niches of species (Stockwell andPeters 1999) and is available at www.lifemap-per.org/desktopgarp. The GARP modelingsystem is a simulation of genetic processes ofmutation, recombination, and selection operat-ing as a nonrandom search for a solution con-sisting of an optimized rule set describing theniche. The data set consists of a series of envi-ronmental grids and species point occurrencedata which are geo-referenced collection sitesfor a particular species but which do not con-sider abundance. GARP analyzes distributioninformation to relate specific ecological char-acteristics to known occurrences. GARP mod-els have been shown to have predictive valuefor mosquitoes that is statistically significant byboth stringent intrinsic tests performed on asubset of the data during development and byextrinsic tests of models using data not in-cluded in model development (Peterson 2001,Levine et al. 2004). We created 10 GARP mod-els and summed them using ArcView GIS 3.3

for each of the final prediction maps. Therefore,individual pixels of all maps were assigned val-ues of 0–10 representing the number of modelsthat predicted occurrence, i.e., ecological suit-ability in that pixel. In the results, we refer tothe number of models in a particular predic-tion map that agree on ecological suitabilitymerely to describe the distribution of pointsand not as a statistical test. Such tests are de-scribed separately. All models were created us-ing a maximum of 100,000 iterations and a con-vergence criterion of 0.0001. The latter criteriondetermines the degree of change between sub-sequent models created during each model de-velopment that is required to continue the it-eration process.

We projected the ecological niche character-istics developed from the mosquito’s nativerange separately onto five other geographiczones: (1) North America and Central America,(2) South America, (3) Europe and the MiddleEast, (4) Australia and the Pacific Islands, and(5) Africa. These projections are produced bymapping areas of each region that are similarto those determined for the native range pre-dictions using the GARP software. Maps cre-ated for each of the five zones were joined to-gether to show a world prediction map for Ae.albopictus establishment.

Subsequent to creating the GARP predic-tion maps based only on ecological suitabil-ity, risk of establishment was determined byconsidering other factors in a stepwise man-ner. As a precondition for risk, countries inwhich ecological conditions similar to thosein the native range were selected. We ob-served that all countries in which establish-ment has occurred have some areas in whichthere was at least weak prediction map agree-ment, therefore we excluded countries nothaving any models agreeing on establishmentfrom further consideration. (Some islands be-low the resolution of the model climate lay-ers are excluded in this analysis.) We in-cluded as infested all countries that have beenreported (Table 1) and five more that—al-though unreported—are likely to be immi-nently infested due to the proximity of geo-referenced collections near or on the countryborder: Peru, Uruguay, Venezuela, Guyana,and French Guiana.

BENEDICT ET AL.78


Among the remaining ecologically suitablecountries, three risk factors were considered: (1)the extent of favorable area for establishment,(2) probable means of introduction via used tireshipments, and (3) shared borders with infestedcountries. Tire exports from infested countriesto uninfested countries for 2002–2004 were ob-tained from the UN Statistics Division COM-TRADE database under the classification HS02-401220. Most data were provided as kilogramper annum, but where data were presented innumber of units, we used the Rubber Manu-facturers Association’s data to produce a con-version factor of 18.337 kg per tire. Data fromall reporters were included, and per annum im-port averages were calculated. Cluster analysiswas performed on the point occurrence precip-itation and temperature variables by multipleregression considering two classes: above andbelow 30 degrees north latitude.

RESULTS AND DISCUSSION

The predicted Asian range map included 202occurrence points of which two fell completelyoutside of the predicted range (Fig. 1): one onthe Island of Hokkaido, Japan, and the otheron the eastern coast of India. Of the remainder,175 data points (87 %) were within pixels onwhich a majority of models (i.e., 6 or more) pre-dicted presence.

A �2 test was performed to determinewhether the distribution of observed occur-rences was random relative to the predicteddistribution. Of the terrestrial pixels used tocreate the Asia prediction (Fig. 1), 72.8%(367,740) of the pixels corresponded to thosewith model agreement of one or greater. There-fore, a random distribution of points wouldpredict that on average, 147 of the occurrencepoints would fall within the predicted range bychance. In fact, 200 of 202 fell within the pre-dicted range (�2 19.1, 1 df, p � 0.0001) confirm-ing that the GARP predictions were not ran-dom. Potential ranges of Ae. albopictus wereprojected elsewhere using the rules and nichedeveloped in this analysis (Fig. 2) (Peterson andVieglais 2001).

Because three qualities of data were avail-able, we discuss three non-Australasian areas,

United States, and Brazil, individually. Data forCentral America (including Mexico), Africa,and Europe and the Middle East were analyzedto determine how many models predicted theoccurrence points. The results for CentralAmerica and Africa were similar: 92% and 86%of the points, respectively, fell within pixelspredicted by a majority of the models, and byfar most were within pixels predicted by 9 or10 models. In contrast, because few pixels inEurope were predicted for presence by a ma-jority of models, only 17 % fell in this range,and 10% fell within pixels for which no mod-els predicted presence, many of these in north-ern Italy.

For the United States, less precise, but moreabundant, occurrence information is available.Two GARP prediction maps (as describedabove), respectively predicted presence in allbut two or 14 of 1011 counties in the UnitedStates in which Ae. albopictus has been reported(Fig. 3). The exceptions were mainly on thewestern fringe of the major predicted distribu-tion. In some cases, lack of sampling and/or re-porting appear to be responsible for the pre-sumed over-prediction: it is highly likely that

FIG. 1. Predicted Australasian range map of Ae. al-bopictus. Darker shades indicate pixels for which highernumbers of models predicted potential suitable nicheswith the darkest shades signifying 10 models. The legendbar shows the 10 colors used. White squares represent theknown occurrence points used to create the models. Yel-low squares are known introduction sites outside of thenative range.

BENEDICT ET AL.80

Arkansas, Mississippi, Alabama, and Virginiahave been extensively colonized by Ae. albopic-tus. Areas that are in fact suitable for Ae. al-bopictus may or may not in fact be occupied ei-ther because of lack of an introduction resultingin establishment, model error, and microcli-mates that support establishment yet fail to berepresented by the relatively coarse resolutionsused in making predictions.

We predict extensive suitable range on thewest coast of the United States that is unoccu-pied. Three possibilities explain this: (1) over-

prediction, (2) introductions resulting in estab-lishment have not yet occurred, or (3) inhos-pitable seasonal conditions with rainy periodscorresponding to cool periods, and warm peri-ods usually being dry. Various authors havepreviously noted this as a potential obstacle to west coast establishment of Ae. albopictus(Nawrocki and Hawley 1987). Egg diapause inthe winter, characteristic of temperate Ae. al-bopictus (Lounibos et al. 2003) would limit hi-bernal development on the West Coast, unlessintroduced populations were of tropical origin,

FIG. 2. Predicted non-Australasian potential distribution of Ae. albopictus. Darker shades indicate greater numbersof models in agreement for suitable habitat with the darkest shades signifying 10 models with the legend showingeach. The predicted distribution in western North America is shown in Figure 3.

FIG. 3. Predicted distribution maps and actual spread of Ae. albopictus in the lower 48 states. The predicted distri-bution areas (red) and the documented spread (yellow) of Ae. albopictus through the year 2001 are shown. One of thetwo prediction maps for the US is shown. Differences between the two consisted largely of one of the ten modelsused to create the prediction map that predicted a broader Texas distribution. Counties colored green are those inwhich introduction has occurred but not establishment.


as observed in southern California in 2001(Madon et al. 2002). While it is within the ca-pabilities of GARP to account for such vari-ables, this was not done in the current analy-sis.

An extensive geo-referenced dataset for es-tablishment of Ae. albopictus in Brazil was avail-able. It is generally consistent with the GARPpredictions (Fig. 4), but because we predict al-most all of Brazil is suitable for Ae. albopictus,over-prediction is not useful to determine theutility of GARP. Of all points, 4334 of 5501 fellwithin pixels in which a majority of modelsagreed on presence.

Two approaches have been used to estimatethe potential range of Ae. albopictus: field dis-tribution information and laboratory observa-tions of the effects of environment on develop-ment and survival. Most emphasis has beenplaced on winter survivorship of temperatepopulations which may be protected by egg diapause (Hawley 1988, Kobayashi et al. 2002,Nawrocki and Hawley 1987). However, theseprovided little detail except for limited areas of Europe (Knudsen 1996, Mitchell 1995), Japan(Kobayashi et al. 2002), and, to a lesser extent,North America (Nawrocki and Hawley 1987).These utilized average January temperatures,degree-days, average annual temperature andprecipitation to circumscribe the potential

range. The �5°C January isotherm, whichNawrocki and Hawley (1987) considered a lib-eral estimate of the northern distribution limit,is similar to that which we predict using GARP,though the latter method resulted in more de-tail and added a prediction of the western limitof potential habitat in the Midwest. Mitchell(1995) suggested that the maximum distribu-tion of Ae. albopictus included limited parts ofSpain, Portugal, France, Italy, Greece, Albania,Macedonia, Bulgaria, and Turkey. However,this model did not predict occupied habitat inboth northern Italy and France predicted byGARP. Knudsen et al. (1996) used a number ofclimate factors to develop a risk map for Euro-pean countries including areas where themonthly mean temperature is 0°C or less inwinter and 20°C or more in summer and withat least 50 cm of mean annual rainfall. Whileascertaining that the maximum range of Ae. al-bopictus was necessarily more extensive thanthe location of the 10°C isotherm, the resultingrisk map classified the nations of Europe intothree categories of high, medium, and low riskand did not attempt to illustrate spatial varia-tion of habitat within each country. Their re-sulting highest risk countries were ultimatelythe same ones presented by Mitchell.

Temperature and humidity values at Ae. al-bopictus occurrence points demonstrate an abil-ity to inhabit either relatively cold and dry orwarm and wet climates (Fig. 5, Table 2), there-fore risk of establishment depends to some ex-tent on the origin of the mosquitoes. Geneticanalyses have corroborated that Ae. albopictusthat colonized the United States were fromtemperate Japan, but colonizing populations ofthe same species in Brazil were of tropical ori-gin (Birungi and Munstermann 2002). In its na-tive range, the distribution of Ae. albopictusextends from Japan and China to tropical coun-tries such as Malaysia and Singapore (Fig. 1;Hawley 1988). Establishment of temperate lat-itudes by this species has been facilitated by anegg diapause which confers cold-hardiness butwhich is absent in tropical populations of Ae.albopictus (Hawley et al. 1987, 1989). However,over time, the incidence of egg diapause hasdecreased as Ae. albopictus has spread intowarmer climates of the southern United States(Lounibos 2002). By contrast, in Brazil, where

FIG. 4. Predicted distribution and actual spread of Ae.albopictus in Brazil as of 2004. Both the predicted distrib-ution (shades of red) and the documented spread (yellowpoints) of Ae. albopictus are displayed.

BENEDICT ET AL.82

Ae. albopictus invaders had no diapause in the1980s (Hawley et al. 1987), this trait was de-tected in a small percentage of populations thathad spread into southern, temperate regions ofthat country by 2001 (Lounibos 2002).

The egg stage of Ae. albopictus suffers highmortality in conditions of low humidity andhigh temperature, and the known interactionbetween these mortality factors (Juliano et al.2002) may account in part for the distributionof U.S. populations of the species in relation torainfall and temperature (Fig. 1) and the largegap in the midcontinental range (Fig. 3). Al-though egg diapause confers some resistanceto desiccation (Sota and Mogi 1992), most U.S.

populations retain a high diapause incidence(Lounibos 2002). Superior tolerance of Ae. ae-gypti eggs to humidity and temperature ex-tremes allows this inferior larval competitor tocoexist with Ae. albopictus or occupy habitatsenvironmentally unfavorable to the Asian tigermosquito (Braks et al. 2003).

The approach we have taken of lumping di-apausing and non-diapausing types means thatour models predict potential distribution in totorather than that of one type or the other. More-over, because we determined potential distrib-ution based on annual climate values, predic-tion of habitat whose seasonal suitabilitywould not be reflected in annual averages

TABLE 2. ANNUAL VALUES FOR CLIMATIC ENVIRONMENTAL VARIABLES AT 408 RANDOM

GEO-REFERENCED POINTS WHERE AE. ALBOPICTUS OCCURS WORLDWIDE

Environmental characteristic Range Mean Median

Mean annual temperature (°C) 5.0–28.5 21.5 23.4Mean maximum annual temperature (°C) 8.2–33.4 26.3 28.55Mean annual minimum temperature (°C) �1.8–24.4 16.7 18.3Mean annual precipitation (cm) 29.2–445.3 169.0 157.0Mean annual wet days (n) 30.0–280.8 167.6 169.2Ground-frost days/month (n) 0–138 14.9 0

FIG. 5. Relationship of average annual precipitation and temperature for a subset of global sites of Ae. albopictus es-tablishments. Note that the data points fall into two clusters of temperature and precipitation (p � 0.0001). Those inthe quadrant beneath approximately 18°C and 225 cm precipitation are located above 30 degrees north latitude (�).For this analysis, all worldwide points plus a subsample of 10 random U.S. counties and 67 early references to oc-currence from Brazil were selected. These are the same points whose values are presented in Table 2.


would be underpredicted or overpredicted.Even with this caveat, previous estimates ofrisk underestimate the potential distribution inpart because annual temperature and rainfalllimit values were more restrictive than therange of values observed for the actual distri-bution of the species. Environmental data forAe. albopictus occurrence demonstrate that itcan establish in areas with much lower annualmean temperatures (5°C–28.5°C ) and less rain-fall (as low as 29 cm annually) than has beensuspected.

Establishment risk from a combination of tireimport introductions and the extent of suitablehabitat was determined. Suitable data wereavailable for 31 uninfested countries (Fig. 6):Mainland Australia, New Zealand, and SouthAfrica are clearly most at risk. Two other hightire import countries—The Netherlands andCanada—have an additional risk factor of bor-dering on infested countries but are sufficientlycold that disease risk is low. Of the Africancountries examined, Guinea, Kenya, Tanzania,Uganda, and Zambia are at greatest risk. Ad-ditional countries for which no tire data were

available also contain extensive areas of suit-able niche: Democratic Republic of Congo,Mozambique, Angola, the Central African Re-public, Ethiopia, Congo, Côte d’ Voire, andGabon. Regardless of whether tires will be themode of introduction, Ae. albopictus dispersionacross the entire continent from the three in-fested countries (Nigeria, Cameroon, andEquatorial Guinea) could occur toward boththe east and west. Only Ethiopia is sufficientlyisolated that introduction may not occur via ei-ther dispersion or tire imports.

The most suitable large areas in Europe oc-cur in France and may be already infested.However, Portugal, the eastern Adriatic coast,eastern Turkey, and the Caspian Sea coast ofRussia are also at risk from both dispersion andtire imports. In the Caribbean and CentralAmerica, Belize, Costa Rica, and Haiti have notreported Aedes albopictus yet all are surroundedby, or adjacent to, infested countries and theythemselves provide suitable niche. Establish-ment seems highly probable if it has not oc-curred already. However, the Bahamas and Ja-maica may yet escape establishment because

FIG. 6. Comparison of tire imports from infested countries and extent of suitable habitat. Tire imports from infestedcountries compared to extent of niche in the 4–10 model agreement range. Not labeled individually within the clus-ter of points in the dotted shape are Germany, Russia, Nepal, Romania, United Kingdom, Benin, Senegal, Pakistan,Guinea, Switzerland, Slovenia, Croatia, Bulgaria, Uganda, Luxembourg, Austria, Macedonia, Hungary, Burkina Faso,Ireland, Zambia, and Malawi.

both are isolated, but no tire importation datawere available.

South America appears to be close to the endof the initial phase of colonization by Ae. al-bopictus. If our presumption of the imminent oractual infestation of five countries borderingBrazil is correct, only three remain uninfested,and all have large areas of suitable niche: Suri-name, Ecuador, and Chile. Only Chile differsand may yet escape invasion because the suit-able area is isolated from dispersion fromneighboring countries. Furthermore, Chile, likesome other countries, has banned the importa-tion of used tires, a measure that will providesome protection. However, free trade consid-erations may win the day over public healthconcerns (www.trade-environment.org/page/theme/tewto/tyrescase.htm).

There are numerous ways in which riskcould be estimated, therefore we present ourmethod as one of many reasonable approaches.By analogy to a forest fire, we based risk on ini-tial introduction—the spark—occurring in fa-vorable ecological habitat—the fuel—with sub-sequent spread dependent upon the proximityand suitability of additional favorable habitat.

In summary, few suitable countries consid-ered have a reasonable prospect of excludingAe. albopictus. Mainland Australia, NewZealand, the Bahamas, Jamaica, Chile andEthiopia are most isolated, but thorough in-spections and control will be necessary to pre-vent infestation.

This study predicts the potential global dis-tribution and risk map for the establishment ofAe. albopictus. Globalization of cargo trade andincreasing air travel means that opportunitiesfor introductions of exotic organisms includingpathogens and their hosts are abundant. Meth-ods used in this analysis have potential for pre-dicting risk from future invasions of vectors orpathogens.

ACKNOWLEDGMENTS

R.S.L. was supported by the APHL/CDCEmerging Infectious Diseases Fellowship Pro-gram. L.P.L. was supported in part by NIHgrant R01 AI-44793 (Invasion Biology of Aedesalbopictus). We thank Chet Moore, Marco

Suarez, and Paul Reiter for valuable directionon data sources. We also appreciate the exten-sive advice in the use of GARP by A. TownsendPeterson. The Brazil dataset was graciouslyprovided by Roseli La Corte dos Santos, FUNASA, Brazil.

REFERENCES

Adhami, J, Reiter, P. Introduction and establishment ofAedes (Stegomyia) albopictus (Skuse) (Diptera: Culici-dae) in Albania. J Am Mosq Control Assoc 1998;14:340–343.

Anonymous. Update: Aedes albopictus infestation—UnitedStates, Mexico. MMWR 1989; 38:445–446.

Birungi, J, Munstermann, LE. Genetic structure of Aedesalbopictus (Diptera: Culicidae) populations based on mi-tochondrial ND5 sequences: Evidence for an indepen-dent invasion into Brazil and United States. Annals En-tomol Soc Am 2002; 95:125–132.

Braks, MAH, Honorio, NA, Lourenco-De-Oliveira, R, Ju-liano, SA, et al. Convergent habitat segregation of Aedesaegypti and Aedes albopictus (Diptera: Culicidae) insoutheastern Brazil and Florida. J Med Entomol 2003;40:785–794.

Broche, RG, Borja, EM. Aedes albopictus in Cuba. J AmMosq Control Assoc 1999; 15:569–570.

Cornel, A, Hunt, RH. Aedes albopictus in Africa? Firstrecords of live specimens in imported tires in CapeTown. J Am Mosq Control Assoc 1991; 7:107–108.

Estrada-Franco, J. Biology, disease relationship, and con-trol of Aedes albopictus. PAHO Technical Paper 1995; 42.

Flacio, E, Lüthy, P, Patocchi, N, Guidotti, F, et al. Primoritrovamento di Aedes albopictus in Svizzera. Bollettinodella Società Ticinese di Scienze Naturli (STSN) 2004;92:141–142.

Fontenille, D, Toto, JC. Aedes (Stegomyia) albopictus(Skuse), a potential new dengue vector in southernCameroon. Emerg Infect Dis 2001; 7:1066–1067.

Forattini, OP. Identificacao de Aedes (Stegomyia) albopic-tus (Skuse) no Brasil. Rev Saude Pub de Sao Paulo 1986;20:244–245.

Hawley, WA. The biology of Aedes albopictus. J Am MosqControl Assoc 1988; 4(Suppl):1–40.

Hawley, WA, Pumpini, CB, Brady, RH, Craig, GBJ.Overwintering survival of Aedes albopictus (Diptera:Culicidae) eggs in Indiana. J Med Entomol 1989;26:122–129.

Hawley, WA, Reiter, P, Copeland, RS, Pumpini, CB, et al.Aedes albopictus in North America: Probable introduc-tion in used tires from northern Asia. Science 1987;236:1114–1116.

Juliano, SA, Lounibos, LP. Ecology of invasive mosqui-toes: Effects on resident species and on human health.Ecology Letters 2005; 8:558–574.

Juliano, SA, O’Meara, GF, Morrill, JR, Cutwa, MM. Des-iccation and thermal tolerance of eggs and the coexis-

BENEDICT ET AL.84


tence of competing mosquitoes. Oecologia 2002;130:458–469.

Knudsen, A, Romi, R, Majori, G. Occurrence and spreadin Italy of Aedes albopictus, with implications for its in-troduction to other parts of Europe. J Am Mosq Con-trol Assoc 1996; 12:177–183.

Kobayashi, M, Nihei, N, Kurihara, T. Analysis of north-ern distribution of Aedes albopictus (Diptera: Culicidae)in Japan by geographical information system. J Med En-tomol 2002; 39:4–11.

Le Maitre, A, Chadee, DD. Arthropods collected from air-craft at Piarco International Airport, Trinidad, West In-dies. J Am Mosq Control Assoc 1983; 43:21–23.

Levine, RS, Peterson, AT, Benedict, MQ. Geographic andecological distributions of the Anopheles gambiae com-plex members predicted using a genetic algorithm. AmJ Trop Med Hyg 2004; 70:105–109.

Lounibos, LP. Invasion by insect vectors of human dis-ease. Ann Rev Entomol 2002; 47L1–2870:233–266.

Lounibos, LP, Escher, RL, Lourenco-de-Oliveira, R.Asymmetric evolution of photoperiodic diapause intemperate and tropical invasive populations of Aedes al-bopictus (Diptera: Culicidae). Ann Entomol Soc Am2003; 96:512–518.

Lugo Edel, C, Moreno, G, Zachariah, MA, López, MM, etal. Identification of Aedes albopictus in urban Nicaragua.J Am Mosq Control Assoc 2005; 21:325–327.

Madon, MB, Mulla, MS, Shaw, MW, Kluh, S, et al. Intro-duction of Aedes albopictus (Skuse) in southern Califor-nia and potential for its establishment. J Vector Ecol2002; 27:149–154.

Mitchell, C. Geographic spread of Aedes albopictus and po-tential for involvement in arbovirus cycles in theMediterranean Basin. J Vector Ecol 1995; 20:44–58.

Nart, P. Aedes albopictus: 1st record from Spain. 2004;PROMED Mail.

Nawrocki, SJ, Hawley, WA. Estimation of the northernlimits of distribution of Aedes albopictus in North Amer-ica. J Am Mosq Control Assoc 1987; 3:314–317.

Ogata, K, Samayoa, AL. Discovery of Aedes albopictus inGuatemala. J Am Mosq Control Assoc 1996; 12:503–506.

Peña, CJ. First report of Aedes (Stegomyia) albopictus(Skuse) from the Dominican Republic. Vector EcolNewsl 1993; 24.

Peterson, AT. Predicting species’ geographic distributionsbased on ecological niche modeling. The Condor 2001;103:599–605.

Peterson, AT, Vieglais, DA. Predicting species invasionsusing ecological niche modeling: New approaches frombioinformatics attack a pressing problem. BioScience2001; 51:363–371.

Reiter, P. Aedes albopictus and the world trade in used tires,1988–1995: The shape of things to come? J Am MosqControl Assoc 1998; 14:83–94.

Reiter, P, Sprenger, D. The used tire trade: A mechanismfor the worldwide dispersal of container breeding mos-quitoes. J Am Mosq Control Assoc 1987; 3:494–501.

Rossi, GC, Pascual, NT, Krsticevic, FJ. First record of Aedesalbopictus (Skuse) from Argentina. J Am Mosq ControlAssoc 1999; 15:422.

Sabatini, A, Raineri, V, Trovato, G . R, Trovato V, ColuzziM. Aedes albopictus in Italia e possible diffusione dellaspecie nell’area mediterranea. Parassitologia 1990;32:301–304.

Savage, HM, Ezike, VI, Nwankwo, ACN, Miller, BR. Firstrecord of breeding populations of Aedes albopictus incontinental Africa: Implications for arboviral transmis-sion. J Am Mosq Control Assoc 1992; 8:101–103.

Schaffner, F, Karch, S. Premiere observation d’Aedes al-bopictus (Skuse, 1894) en France metropolitaine. AcadSci Paris Sciences de la Vie 2000; 323:373–375.

Schaffner, F, Van Borstel, W, Coosemans, M. First recordof Aedes (Stegomyia) albopictusi in Belgium. J Am MosqControl Assoc 2004; 20:201–203.

Shroyer, DA. Aedes albopictus and arboviruses: A concisereview of the literature. J Am Mosq Control Assoc 1986;2:424–428.

Sota, T, Mogi, M. Survival-time and resistance to desic-cation of diapause and nondiapause eggs of tempera-ture Aedes (Stegomyia) mosquitos. Entomolexpappl1992; 63:155–161.

Sprenger, D, Wuithiranyagool, T. The discovery and dis-tribution of Aedes albopictus (Skuse) in Harris Country,Texas. J Am Mosq Control Assoc 1986; 2:217–219.

Stockwell, DRB, Peters, D. The GARP modelling system:Problems and solutions to automated spatial predic-tion. Int J Geog Inf Sci 1999; 13:143–158.

Tabachnick, WJ. Evolutionary genetics of arthropod-borne diseases. The yellow fever mosquito. Am Ento-mologist 1991; 37:14–24.

Toto, JC, Abaga, S, Carnevale, P, Simard, F. First reportof the oriental mosquito Aedes albopictus on the WestAfrican island of Bioko, Equatorial Guinea. Med VetEntomol 2003; 17:343–346.

Velez, Iñ, Quiñones, ML, Suarez, M, Olano, V, et al. Pres-encia de Aedes albopictus en Leticia, Amazonas, Colom-bia. Biomedica 1998; 18:182–198.

Address reprint requests to:Mark Q. Benedict

Centers for Disease Control and Prevention4770 Buford Highway

Mailstop F-42Atlanta, GA 30341

E-mail: [email protected]

spread of the tiger: global risk of invasion by the...

Documents