s ,d temporal distribution of aedes aegypti in different...

SAMPLING, DISTRIBUTION, DISPERSAL

Temporal Distribution of Aedes aegypti in Different Districts of RioDe Janeiro, Brazil, Measured by Two Types of Traps

N. A. HONORIO,1,2 C. T. CODECO,3 F. C. ALVES,4 M.A.F.M. MAGALHAES,5

AND R. LOURENCO-DE-OLIVEIRA1

J. Med. Entomol. 46(5): 1001Ð1014 (2009)

ABSTRACT Dengue dynamics in Rio de Janeiro, Brazil, as in many dengue-endemic regions of theworld, is seasonal, with peaks during the wetÐhot months. This temporal pattern is generally attributedto the dynamics of its mosquito vector Aedes aegypti (L.). The objectives of this study were tocharacterize the temporal pattern of Ae. aegypti population dynamics in three neighborhoods of Riode Janeiro and its association with local meteorological variables; and to compare positivity and densityindices obtained with ovitraps and MosquiTraps. The three neighborhoods are distinct in vegetationcoverage, sanitation, water supply, and urbanization. Mosquito sampling was carried out weekly, fromSeptember 2006 to March 2008, a period during which large dengue epidemics occurred in the city.Our results show peaks of oviposition in early summer 2007 and late summer 2008, detected by bothtraps. The ovitrap provided a more sensitive index than MosquiTrap. The MosquiTrap detectionthreshold showed high variation among areas, corresponding to a mean egg density of �25Ð52 eggsper ovitrap. Both temperature and rainfall were signiÞcantly related to Ae. aegypti indices at a short(1 wk) time lag. Our results suggest that mean weekly temperature above 22Ð24�C is stronglyassociated with high Ae. aegypti abundance and consequently with an increased risk of denguetransmission. Understanding the effects of meteorological variables on Ae. aegypti population dy-namics will help to target control measures at the times when vector populations are greatest,contributing to the development of climate-based control and surveillance measures for dengue feverin a hyperendemic area.

KEY WORDS population dynamics, Aedes aegypti, meteorological variables, dengue, traps

Dengue is the most prevalent mosquito-borne viraldisease in the world and an increasingly importantcause of morbidity and mortality. Since the reinvasionof Brazil by the primary dengue vector Aedes aegypti(L.) in 1977, this country has experienced severaldengue epidemics, with 4.5 million cases of dengueuntil 2008 (Ministerio da Saude 2008). Dengue wasÞrst recognized in Brazil in 1981Ð1982, during an out-break in the city of Boa Vista, Roraima, northern Bra-zil. In this outbreak, DENV-1 and DENV-4 serotypeswere isolated, and 11,000 cases were conÞrmed(Osanai et al. 1983). Countrywide transmission ofDENV-1 began only 4 yr later, when a large dengueepidemic started in metropolitan Rio de Janeiro. Riode Janeiro was the port of entry for DENV-2 andDENV-3 into the country, in 1990 and 2001, respec-tively. In 2001Ð2002, DENV-3 caused a widespread

and severe epidemic with 368,460 cases and 91 deaths(Schatzmayr 2000; Nogueira et al. 2001, 2002). In2007Ð2008, during this study, an even more severeepidemic (in terms of case-fatality ratio) occurred(Lourenco-de-Oliveira 2008), with 316,287 cases no-tiÞed until October 2008, in Rio de Janeiro, and 174deaths (52 from dengue hemorrhagic fever, 39 casesfrom dengue shock syndrome, and 83 from other den-gue-related complications [SESDEC/RJ 2008]). Thislatter epidemic, mainly caused by DENV-2, was char-acterized by a shift toward lower age groups (Teixeiraet al. 2008).

Rio de Janeiro is the second largest city of Brazil,with �6 million inhabitants (IBGE 2000). Denguedynamic in Rio de Janeiro, as in many endemic regionsof the world, is seasonal, with the highest dengueactivity during the late summer (Nogueira et al. 2002,Vezzani and Carbajo 2008, revised by Halstead 2008).Monthly mean temperature and rainfall in Rio deJaneiro (10-yr average) show that the area is charac-terized by a rainy warm season (NovemberÐApril),with mean monthly temperatures ranging between 24and 27�C and rainfall ranging between 100 and 170mm/mo; and a dry cool season (MayÐOctober), withmean temperatures ranging between 21 and 24�C andrainfall between 50 and 100 mm/mo (http://www.

1 Laboratorio de Transmissores de Hematozoarios, Instituto Os-waldoCruz-Fiocruz,Av.Brasil 4365,Manguinhos,Riode Janeiro,CEP21045-900, Brasil.

2 Corresponding author, e-mail: honorio@ioc.Þocruz.br.3 Programa de Computacao CientõÞca-Fiocruz, Rio de Janeiro, RJ,

Brasil.4 Secretaria Municipal de Saude do Rio de Janeiro-Fundacao Na-

cional de Saude, Av. Pedro II 278, Sao Cristovao, Rio de Janeiro, Brasil.5 Instituto de Comunicacao e Informacao CientõÞca e Tecnologica

em SaudeÐICICT-Fiocruz, Rio de Janeiro, RJ, Brasil.

0022-2585/09/1001Ð1014$04.00/0 � 2009 Entomological Society of America

bdclima.cnpm.embrapa.br/). The seasonal dynamicsof dengue transmission has been attributed to theeffect of climate on its vector population dynamics.Temperature and rainfall inßuence many aspects ofAe. aegypti life history: temperature affects egg hatch(Gubler 1988; Focks et al. 1993a,b), egg viability(Parker 1986), larval development (Rueda et al. 1990),blood-feeding behavior (Crans et al. 1996), length ofthe gonotrophic cycle (Pant and Yasuno 1973), femalefecundity (Day et al. 1990), adult size (Rueda et al.1990, Tun-Lin et al. 2000), and adult longevity anddispersal (Hawley 1985, Honorio et al. 2003, Maciel-de-Freitas et al. 2007a). Rainfall may affect the abun-dance and productivity of breeding sites by reducinglarval competition and stimulating egg hatching(Moore et al. 1978, Lounibos 1981, Honorio et al. 2006,Maciel-de-Freitas et al. 2007b). Low relative humidityhas been associated with reduced egg viability, adultsurvival, and fecundity (Canyon et al. 1999, Costanzoet al. 2006, Luz et al. 2008).

Dengue transmission and vector distribution alsoshow strong spatial heterogeneity. In Rio de Janeiro,higher infestation levels are associated with denselypopulated neighborhoods, in which unplanned urban-ization, irregular sanitation and water supply, and lackof garbage collection foster proliferation of potentialbreeding sites forAe. aegypti, affecting the abundanceofdenguevectors and transmissionof thedenguevirus(Tauil 2001, Luz et al. 2003).

Mathematical models suggest that variability inmosquito recruitment, together with short-term crossimmunity to virus serotypes, are important factorsmodulating dengue temporal dynamics as well as thecoexistence of multiple strains (Bartley et al. 2002,Wearing and Rohani 2006). Thus, spatial and temporalvariation in climate, along with spatial variation inhuman population and conditions in the human en-vironment, should affect conditions for the growth ofthe primary dengue vector (Wu et al. 2007, Melo et al.2007) and the transmission of dengue viruses. Despitethe importance of climate for the transmission of den-gue, few studies have examined detailed longitudinaldata sets describing Ae. aegypti population dynamics(Scott et al. 2000).

In this study, we present the results of a 1.5-yrcontinuous survey ofAe. aegyptipopulation dynamics,encompassing two hot-wet seasons and one dry-coolseason, in three distinct neighborhoods of Rio de Ja-neiro, by using two types of traps that attract ovipos-iting females. These neighborhoods differ in income,urbanization, access to water (which inßuences waterstorage habits) and garbage collection, presence ofyards, vegetation coverage, and history of dengueprevalence. The main objective was to characterizethe temporal distribution of Ae. aegypti abundance inRio de Janeiro and its association with local meteo-rological variables. We hypothesize that temperatureshould have a similar effect in all three areas, whereasthe effect of rainfall should vary according to charac-teristics of the areas, such as size of outdoor areas andavailability of more productive containers.

Materials and Methods

Study Areas. Surveys were performed in threeneighborhoods of Rio de Janeiro city: Higienopolis,Tubiacanga, and Palmares, which differ in human pop-ulation density, sanitation, vegetation cover, and his-tory of dengue (Table 1; Fig. 1). Higienopolis andTubiacanga are geographically closer (13.1 km apart)than Palmares (30.7 km and 45.1 km from Higienopolisand Tubiacanga, respectively). In March 2007, houseindex (percent houses infested by Ae. aegypti) (Con-nor and Monroe 1923) was 3.03, 9.83, and 19.48 forPalmares, Higienopolis, and Tubiacanga, respectively.In 2008, Rio de Janeiro city notiÞed 125,988 denguecases and 100 deaths. In this epidemic, the district ofHigienopolis reported 350 dengue cases, whereas 583were reported in the Galeao area (where the neigh-borhood of Tubiacanga is located) and 580 in theVargem Pequena area (where the suburban slum ofPalmares is located). The total incidence rates (1:100,000) were 2110.1, 2695.0 and 4802.4 in Higienopo-lis, Galeao, and Vargem Pequena, respectively(SMS/RJ 2008).Higienopolis. Higienopolis (22� 52�25� S, 43� 15�41�

W) is a neighborhood located within a densely pop-ulated urban area in Rio de Janeiro city (human pop-

Table 1. Characteristics of the three study areas based on Census data and surveyed houses

CharacteristicHigienopolis(urban), %

Tubiacanga(suburban), %

Palmares(suburban slum), %

P (�2 test)a

Access to piped waterb 99.3 99.1 29.3 �0.001Only artesian wellb 0.0 14.0 64.4 �0.001Sewage serviceb 92.2 46.6 24.2 �0.001Garbage serviceb 99.3 99.4 97.5 0.89Poorly maintained householdsc 4.9 25.9 20.0 �0.001Households with yards (with trees/ cemented)c 33.3/56.8 44.5/45.7 52.5/32.5 0.18/0.07Reporting daily access to waterc 92.6 85.2 15.0 �0.001Reporting no piped waterc 0.0 0.0 45.0Yards with potential larval habitatsc 48.1 54.3 63.8 0.4Head of family with only basic schoolingc 34.6 23.5 27.5 0.41Graduated head of familyc 25.9 3.7 5.0 �0.001

aChi-square test for differences among areas.bCensus data (IBGE 2000).c Surveyed houses (n � 80 per neighborhood).

1002 JOURNAL OF MEDICAL ENTOMOLOGY Vol. 46, no. 5

ulation density 15,891 inhabitants per km2). One of themajor cityÕs highways, the Yellow Line, crosses theregion. The area is totally urbanized, streets are paved;water supply and garbage services are good. Residents,mostly middle class and aged, live in houses with no orsmall cemented yards. Vegetation coverage is low.This neighborhood, however, is surrounded by someof the largest slums of Rio de Janeiro, with favorableconditions for Ae. aegypti proliferation.Tubiacanga. Tubiacanga (22� 47�08� S, 43� 13�36�

W) is a small suburban neighborhood located at theborder of the Governador Island, in the GuanabaraBay (human population density 8,219 inhabitants perkm2). This mainly residential neighborhood is mostlyoccupied by houses with large open yards. Streets areunpaved, garbage collection is regular, but access towater is irregular, and residents often store water incontainers, which are potential development sites forimmature Ae. aegypti (Maciel-de-Freitas et al.2007a,b).Palmares. Palmares (22� 59�26� S, 43� 27�36� W) is a

disordered human settlement started in the 1990s, asuburban slum located in one of the major axes ofexpansion of the city, to the West. Between the rainforest mountainous range and a polluted river, thehuman population density in Palmares is 2,732 inhab-itants per km2. Housing distribution is crowded andirregular, with narrow unpaved alleys. Sixty-six per-cent of the houses investigated reported obtainingwater from wells, and water storage in containers iscommon. Vegetation coverage is low within the neigh-borhood but very dense around it.

Meteorological Variables. Air temperature wasmeasured at two meteorological stations located �5km from each of the study areas (the same tempera-ture data were used for Tubiacanga and Higienopolis).Rainfall data for each area were obtained from Prefei-tura da Cidade do Rio de Janeiro (GeoRio-http//www.rio.rj.gov.br/georio/alerta/tempo).Entomological Survey. Two types of traps for ovi-

positing females were used, ovitraps and MosquiTraps.Ovitraps are black plastic containers, Þlled with 300 mlof a10%hay infusion, andawoodenpaddleheldon thewall for oviposition (Fay and Eliason 1966, Reiter et al.1991, Honorio et al. 2003). The MosquiTrap version 1.0(Ecovec Ltd., Belo Horizonte, MG, Brazil) consists ofa matte black container (16 cm in height by 11 cm indiameter) with �280 ml of water and a removablesticky card. A synthetic oviposition attractant, AtrAedes(Ecovec Ltd.), is placed inside the MosquiTrap toattract gravid female mosquitoes (Favaro et al. 2006).Thus, both traps attract gravid females, but the ovitrapcaptures the eggs, whereas the MosquiTrap capturesthe ovipositing female. Using two types of traps for thesame population enabled us to generate two simulta-neous, independent, time series for each area, whichwill be used as replicates for validating the estimatedmeteorological variables effects.

The entomological survey was carried out weekly,from September 2006 to March 2008 during 82 con-tinuous weeks, encompassing two wet-hot seasons andone dry-cool season. In each area, a grid of 500 by500 m was deÞned and 80 points randomly selected.The house nearest to each point was selected. Only

Fig. 1. Map of Rio de Janeiro, Brazil, showing the location of the three studied neighborhoods.

September 2009 HONORIO ET AL.: TEMPORAL DISTRIBUTION OF Ae. aegypti IN RIO DE JANEIRO 1003

houses �50 m apart were considered. Householderswere invited to participate in the study and, uponconsent, were interviewed. We recorded informationabout the house characteristics and the number andeducation of household members (Table 1). AfterownerÕs written consent, traps were placed in these 80randomly selected houses. One ovitrap was placed ineach of 40 houses and one MosquiTrap was placed ineach of the other 40 houses. Traps were placed in theperidomestic area, preferably in the garden or in theshade. All traps were inspected weekly during 82 wk.During the inspection, hay infusion in the ovitraps wasreplaced and wooden paddles collected and individ-ually packed in plastic bags. In the laboratory, positivepaddles were stored at 25Ð28�C and �80% humidityand immersed in water for 1 wk to hatch larvae.Hatched larvae were reared to fourth instar and iden-tiÞed. MosquiTraps, during the inspections, had theirsticky cards taken to the laboratory where capturedmosquitoes were removed with the aid of probes andforceps, to be identiÞed (Consoli and Lourenco-de-Oliveira 1994 taxonomic keys) and counted under astereomicroscope. Attractants and sticky cards wereusually replaced every 4 wk, but sticky cards werereplaced earlier when found to be dusty or dirty. Theset of inspected 80 houses in each area remained thesame during the whole study (Table 1).Data Analysis. For the analysis, we calculated two

indices: the proportions of positive ovitraps andMosquiTrapsperweek(“positive”meansat leastoneeggor adult ofAe. aegypti) and density indices, calculatedas the total number of eggs (or adults) collected in anarea, per week, divided by number of traps. To test fordifferences between areas in total number of Ae. ae-gypti captured, we used chi-square tests.

To infer associations between meteorological vari-ables and mosquito infestation, we Þt generalized lin-ear models, using negative binomial distribution and alog link function. The response variable was either theabundance of eggs per week, (Dw), or the abundanceof adults per week, (Aw). The explanatory variableswere lagged mean air temperature and lagged totalprecipitation (mm rainfall). An investigation of auto-correlation plots suggested that an auto-regressiveterm of Þrst order was also necessary.

The full model took the following form:

Log[E(Yw)] � a0 � a1 Yw � 1 b0 Tempw

� b1 Tempw � 1 b2 Tempw � 2 � . . .

� bn Tempw � n � c0 Rainw � c1 Rainw � 1

� c2 Rainw � 2 � . . . � cn Rainw � n et

where E(Yw) is the expected mosquito abundance(either Dw or Aw), Tempw is the mean air temperatureat week w, and Rainw is the total amount of rainfall atweek w (millimeters). The indices w-1 to w-n standsfor the time lags.

Because there is strong correlation between tem-perature and rainfall measured a few weeks apart,multicolinearity is a problem when Þtting such model,which may lead to unstable estimation. To reduce this

problem, we opted to use distributed lag modeling(DLM), an approach that has been used in socialsciences and pollution epidemiology to estimatelagged effects of climate more efÞciently (Souza et al.2007). In distributed lag models, the variation of thecoefÞcients bi (i are the lags) is constrained to Þt apolynomial function. The motivation behind this ap-proach is the realization that the effect of temperatureand rainfall on mosquito density is not instantaneous,but distributes through time in a smooth way.

The same set of nested models was Þtted to all threelocalities: an AR1 model, a model with lagged tem-perature only, and a model with lagged temperatureplus lagged rainfall (we also Þtted models with rainfallonly, which resulted in poorer Þtting compared withtemperature only). AkaikeÕs Information Criterion(AIC) was used to assess Þtness improvement due tovariable addition (Akaike 1974).

To investigate potential nonlinearities on the effectsof rainfall and temperature on abundance indices, wealso Þtted a set of generalized additive models (GAM)with log link function and the negative binomial dis-tribution, to the same outcome variables. GAM is anextension of generalized linear models that allows forthe inclusion of nonparametric smoothing terms(Wood 2006) in the place of the constant parameters.By plotting the Þtted smooth terms versus the predic-tor, one may investigate the nature of the relationshipbetween the predictor and the outcome variable, de-tecting potential nonlinearities.

We started with models with the same structureshown for the DLM analysis, but only keeping thetemperature and rainfall lagged terms that provedsigniÞcant, according to this analysis. Models with andwithout smoothing terms were compared using AIC(Akaike 1974). Models were Þtted in R 2.6.0, using thelibraries mgcv and MASS (R Development Core Team2006). Goodness-of-Þt was assessed by AIC, residualplots, autocorrelation function plots, and SpearmanÕscorrelation between predicted and observed values.

Results

General Entomological Results. Using ovitraps,703,649 eggs of Aedes species were collected, the ma-jority being Ae. aegypti (Table 2) and the remainingbeingAedes albopictus (Skuse). Approximately half ofall eggs were collected in the suburban area (Tubia-canga), one third in the urban area (Higienopolis),and the remaining 18% in the suburban slum (Pal-mares). Differences in total eggs collected per areawere signiÞcant (�2 � 84128, df � 2, P� 0.001) (Table2).

Using MosquiTraps, 15,059 mosquitoes were cap-tured. Contrasting with the ovitrap, MosquiTraps alsoattracted non Aedes species, mainly Culex quinquefas-ciatus Say. This was especially noticeable in the sub-urban slum (which is bordered by a polluted river),where �85% of all specimens collected were non-Aedes species (Table 2). Comparing the three areas,traps in the suburban slum capturedAe. aegypti adultsthe least (only 7% of all captures), whereas traps in the


other two neighborhoods shared similar values (�2 �0.0034, df � 1, P� 0.9533). Ninety-eight percent of allAe. aegypti collected were females.Temporal Pattern of Positive Premises. We found

very high proportion of positive ovitraps per week inboth urban and suburban neighborhoods (Higienopo-

lis and Tubiacanga, respectively), with indices aver-aging 70Ð80% year-round and summer peaks reaching90Ð100%. In the winter, infestation declined but rarelydropped below 60% (Fig. 2). The suburban slum, how-ever, had lower infestation, with an average of 50%positive premises year round, summer peaks of 70Ð

Table 2. Descriptive statistics of meteorological variables and infestation indices based on 40 MosquiTraps and 40 Ovitraps installedfor 82 wk in three study areas in Rio de Janeiro, Brazil

Higienopolis(urban)

Tubiacanga(suburban)

Palmares(suburban slum)

P (�2 test)a

MosquiTrapMosquitoes 4,571 5,247 5,241 �0.001Ae. aegypti 3,977 (87%) 3,854 (73%) 641 (12%) �0.001Mean SD 1.2 0.4 1.2 0.5 0.2 0.1% female Ae. aegypti 98 98 96Ae. albopictus 40 (0.8%) 173 (3.3%) 131 (2.5%) �0.001% female Ae. albopictus 95 100 92% other mosquitoes 12 24 85 �0.001% Missing traps 3 1.7 2.7

OvitrapNumber of eggs 252,818 (36%) 323,476 (46%) 127,355 (18%) �0.001% Ae. aegypti 98 96 91Mean SD 77 30 101 44 39.7 20% Ae. albopictus 0.1 0.4 2.2% Missing traps 3 1.6 2.9

WeatherRainfall (mm) 1392 1584 1997 �0.001% Rainy daysb 34 34.9 35.9Mean temp (�C) 24.4c 24.4c 25.6

aChi-square test for differences among areas.b Rainy day was any day with rainfall (in millimeters) � 0.cData obtained at the same meteorological station.

Fig. 2. Time series of mean temperature (Celsius) per week, total rainfall (millimeters), positive MosquiTraps (per-centage), and positive ovitraps (percentage), from September 2006 to March 2008, in three neighborhoods of Rio de Janeiro,Brazil: Higienopolis (urban), Tubiacanga (suburban), and Palmares (suburban slum). The solid curve is a natural cubicsmoothing spline, and the horizontal line indicates the overall mean trap positivity.


90%, and winter decline to 20Ð30%. Time series of allthree areas presented oscillations and the second sum-mer had the highest peak.

Using the proportion of positive MosquiTraps tocalculate infestation, we obtained lower indices rela-tive to ovitraps. Again, the suburban slum was the leastinfested, with an average of 15% positivity, summerpeaks of �30%, and winter declines to 5Ð10%. Theurban and suburban areas shared similar infestationlevels (�50%; summer peaks of �70%, and winterdeclines to 30Ð40%) (Fig. 2).

Both traps produced the same ranking of areas,though the summer peak was less evident with theMosquiTrap data. Only data from the suburban slumshowed the two summer peaks, whereas the suburbanarea showed only the Þrst peak. There was no sug-gestion of temporal pattern in the urban area positivitydata. (Fig. 2).Temporal Pattern of Density Indices. Using ovit-

raps, we calculated the egg density index per week(Fig. 3). With this index, we were able to discriminatebetween suburban and urban areas, whose valuesseemed similar when comparing the positivity indices.The suburban area emerges as the most productivearea, with a mean of 100 eggs per trap per wk, whichdoubled during the summer peaks and dropped to halfduring the winter. The urban area was second mostproductive, averaging �77 eggs per trap per wk, dou-bling during the summer. The suburban slum was theleast productive, with a mean of 39.7 eggs per trap perwk, and summer peaks of 60Ð70 eggs per trap per wk.

The summer peak is evident in suburban and suburbanslum and less clear in urban area (no peak in thesecond summer).

Using MosquiTraps, we calculated the adult densityindex (Fig. 3). This index did not discriminate be-tween the urban and suburban areas, both with anaverage of �1.3 mosquitoes per trap wk. The suburbanslum presented very low indices, averaging only 0.3mosquitoes per trap per wk. The summer peak is againevident in the suburban area and the suburban slum.Trap Sensitivity and Comparison Between Indices.

Ovitraps were more sensitive than MosquiTraps fordetecting Ae. aegypti presence as well as measuringinfestation intensity. Evidence for this conclusioncomes from the higher indices obtained with ovitraps(Fig. 3) and from scatter plots of the total number ofeggs versus total number of adults collected per week,in each area (Fig. 4). Fitting linear regressions of theformDw � b0 b1 Aw, we obtained estimates of b0 (ameasure of sensitivity of the ovitraps), and b1 (a con-version factor adults-eggs) (Table 3). The MosquiTrapdetection threshold showed high variation among areas,corresponding to a mean egg density of �25Ð52 eggsper ovitrap, that is, below this amount of eggs, Mos-quiTraps did not detect the presence of Ae. aegypti.For egg density indices above this threshold, eachcaptured adult corresponded to an average of 41.9 9.6 eggs per adult in the suburban area (Tubiacanga),52.8 9.3 eggs per adult in the urban (Higienopolis),and 24.2 3.6 eggs per adult in the suburban slum(Palmares) (Table 3), indicating high variation in de-

Fig. 3. Time series of mean temperature (Celsius) per week, total rainfall (millimeters), mean adult density perMosquiTraps, and mean egg density per ovitraps from September 2006 to March 2008, in three neighborhoods of Rio deJaneiro, Brazil: Higienopolis (urban), Tubiacanga (suburban), and Palmares (suburban slum).


tection among areas. These numbers should be con-sidered with caution because differences between de-tection thresholds may be artiÞcially caused byassuming linearity, and the lack of data on small in-festation range in two of the areas.

In two neighborhoods (suburban and suburbanslum) the two traps produced moderately correlateddensity indices (Tubiacanga (suburban): SpearmanÕsr� 0.60, P� 0.001; Palmares (suburban slum): Spear-manÕs r � 0.49, P � 0.001). In the remaining area,however, only a weak correlation was observed(SpearmanÕs r � 0.29, P � 0.01) (Table 3).Meteorological Variables. The suburban slum, lo-

cated at the border of the Rio de JaneiroÕs rain forestmountain range, received the greatest precipitation,with a mean of 98 mm rainfall per mo and well-dis-tributed strong precipitation events throughout theyear (Fig. 2). In contrast, the urban and suburbanareas, which are located in the rain-shadow of thecoastal mountain range, shared lower average precip-itation (average of 68Ð78 mm rainfall per month), withintense events more concentrated in the summer.Meteorological Variables and Infestation. Autocor-relation. We started the regression modeling by in-specting the autocorrelation structure of the egg andadult time series. In two areas (suburban slum and

suburban area), both egg and adult time series pre-sented signiÞcant autocorrelation, at least at lag 1(adult series presented longer autocorrelated struc-ture). In contrast, the urban area presented no evi-dence of autocorrelation. A simple AR model of Þrstorder was sufÞcient to remove the autocorrelationfrom all series (data not shown).Lag Distribution of Climate Effects. Figure 5 shows

the estimated lag distributed effect of temperatureand rainfall on egg and adult mosquito abundance. Inall series, the estimated lagged effect of temperatureshowed a strong positive effect at lag 1. Despite fol-lowing the same pattern, suburban slum showed lesssigniÞcant effects, possibly because of the lower in-festation levels (which reduced the statistical power).Including lag distributed mean temperature in theAR1 model signiÞcantly improved the goodness-of-Þtin both egg and adult response variables in all threeareas (Table 4).

The estimated lagged effect of rainfall was onlysigniÞcant in the suburban area, where at lag 3, bothegg and adult time series indicate a signiÞcant positiveeffect. At lag 1, both series suggest a negative effect,but only the egg time series was signiÞcant. In theurban area, there is some indication of a positive effectat lags 4 or 5. In the suburban slum, the same is found,

Fig. 4. Scatterplots of number of eggs per ovitrap versus number of adults per MosquiTrap captured per week, in threeneighborhoods of Rio de Janeiro, Brazil: Higienopolis (urban), Tubiacanga (suburban), and Palmares (suburban slum). Solidline: linear regression. Dotted line, spline.

Table 3. Result of fitting the linear model Egg � b0 � b1 Adult in the three study areas of Rio de Janeiro

Higienopolis (urban) Tubiacanga (suburban)Palmares

(suburban slum)

Intercept 52.8 9.3 eggs per adult 41.9 9.6 egg per adult 24.2 3.6 egg per adultSlope 19.5 7 egg per adult 49 7 egg per adult 77 15 egg per adultPearsonÕs r 0.29** 0.60*** 0.49***

** P � 0.05; *** P � 0.001.


but only in the egg time series. Only the egg time seriesmodels for the suburban and the suburban slum wereimproved by the inclusion of lagged rainfall in a modelwith AR1 and temperature.

The effect of temperature at lag one was the onlyterm that consistently presented a nonlinear pattern,the only exception being the adult time series in the

urban area (Fig. 6). The pattern obtained suggests thatthe effect of temperature at lag 1 increases linearlyuntil a threshold (�22Ð24�C), above which its effectbecomes less signiÞcant.

Besides temperature at lag 1, all other effects wereadequately represented by Þxed terms. The only ex-ception was rainfall at lag 4, in the suburban area, but

Fig. 5. Lag distributed effect of temperature and rainfall on (A) the mean abundance of eggs, and (B) adults ofAe. aegyptiin three neighborhoods of Rio de Janeiro, Brazil: Higienopolis (urban), Tubiacanga (suburban), and Palmares (suburbanslum). Dotted lines indicated 95% conÞdence interval.

Table 4. Comparison of nested DLM models

Response variable ModelHigienopolis (urban) Tubiacanga (suburban)

Palmares (suburbanslum)

r AIC p (�2) r AIC p (�2) r AIC P (�2)

Mean egg density Null 736 776 658AR 0.24 723 �0.001 0.48 738 �0.001 0.36 639 �0.001AR temp 0.48 713 0.001 0.58 736 0.038 0.48 638 0.053AR temp rainfall 0.57 714 0.123 0.69 733 0.022 0.62 635 0.025

Adult capture rate Null 665 670 451AR 0.12 657 0.007 0.46 649 �0.001 0.57 415 �0.001AR temp 0.47 651 0.008 0.60 644 0.015 0.59 421 0.58AR temp rainfall 0.50 655 0.295 0.65 646 0.155 0.59 429 0.82

r, Spearman correlation coefÞcient; AIC, Akaike information criterion.


the pattern found was inconclusive, as conÞdenceintervals were very large (data not shown). Table 5presents the results of the best Þtted GAM models.Goodness-of-Fit. The best models (either DLM

models or GAM) predicted time series that matchedthe observed series reasonably well, with SpearmanÕscorrelation coefÞcients ranging �0.6 (Tables 4 and 5).Figure 7 presents the time series of Þtted and observedegg/adult density. Peaks in the observed data werecaptured by the models several times, but some ofthem were also missed badly. There are two peaks inthe Þrst summer of the suburban area, which werecaptured very well in the egg time series, but not in theadult time series. In the urban area, which has a noisier

data, most peaks are missed. In the suburban slum,Þtting is better for the adult time series. Residual plotsshow that both DLM and GAM models captured theoverall structure of the original data (data not shown).

Discussion

We investigated the temporal dynamics of Ae. ae-gypti abundance indices in three neighborhoods ofRio de Janeiro by using ovitraps and MosquiTraps.Ovitraps are standard tools for Ae. aegypti populationsampling in many settings, considered a sensitive andcost-effective surveillance apparatus (Fay and Eliason1966, Braga et al. 2000). They are defended as an

Fig. 6. Smooth effect of temperature at lag 1 wk on the abundance of Ae. aegypti adults and eggs: Higienopolis (urban),Tubiacanga (suburban), and Palmares (suburban slum). Dotted lines indicated 95% conÞdence interval.

Table 5. Best generalized additive models for egg and adult Ae. aegypti density (lag measured in weeks)

Area Variable Lag Effect SE P value AIC r

Higienopolis Egg 1 0.002 0.001 * 706 0.61(urban) Temp 1 Smooth ***

3 �0.050 0.019 **Rain 4 Smooth **Adult 1 0.001 0.002 n.s 743 0.43Temp 1 0.073 0.019 ***

3 �0.055 0.019 ***Tubiacanga Egg 1 0.005 0.100 *** 723 0.66(suburban) Temp 1 Smooth *

Rain 1 �0.004 0.001 ***3 0.003 0.001 **

Adult 1 0.005 0.002 *** 638 0.57Temp 1 Smooth ***

3 �0.04 0.020 *Palmares Egg 1 0.006 0.002 ** 629 0.53(slum) Temp 1 Smooth **

5 �0.046 0.020 **Rain 4 0.002 0.001 *Adult 1 0.061 0.010 *** 410 0.62Temp 1 Smooth *

Chi-square partial testÕs P value: * P � 0.1, ** P � 0.05, *** P � 0.001.AIC, Akaike information criterion; r, Spearman correlation between observed and Þtted values.


excellent tool to detect the presence of the vector andto compare infestations among different areas (Focks2003, Vezzani et al. 2004, Morato et al. 2005, Rõos-Velasquez et al. 2007, Regis et al. 2008). However, theyare considered poor indicators of transmission riskbecause their measurement of adult population den-sity is indirect (Focks 2003). Sticky ovitraps such asthe MosquiTrap (Eiras 2002, Ritchie et al. 2004, Fac-chinelli et al. 2007); however, are advocated as a betterapproach, as they would provide direct measurementsof segment of the population (adult females) involvedin transmission (Focks 2003, Eiras 2002, Gama et al.2007, Favaro et al. 2008).

Despite the main difference (one capturing eggsthe other capturing ovipositing females), we proposedthat, because both traps target the same population(ovipositing females), their results should be compa-rable. Our results seem to support this hypothesis.First, both egg and adult time series produced quali-tatively similar time series of infestation in the sub-urban area and suburban slum, with both series show-ing moderate linear correlation (Fig. 2). In these samelocalities, both traps also detected the summer peaksand winter declines. The time series obtained for thedensely populated urban neighborhood (Higienopo-lis), however, was noisier and less structured, and theovitrap and MosquiTrap data showed no correlation.These results may reßect the inßuence of immigrationofgravid females fromsurrounding sites into theurbanneighborhood. In contrast to this urban neighbor-hood, the suburban slum and the suburban area areboth geographically isolated areas (Lourenco-de-Ol-iveira et al. 2004). When comparing the three areas,both traps also tended to agree, producing the sameranking of infested areas (suburban and urban versussuburban slum). Finally, statistical analysis of the ovit-rap and MosquiTrap time series tended to agree on the

effect of meteorological variables onAe. aegypti abun-dance.

Ovitraps showed greater sensitivity than Mosqui-Traps when used for detecting the presence of mos-quitoes in households. Ovitraps always providedhigher percentage of positive premises (Fig. 2). Linearregression (Table 4) suggests that the relationshipbetween MosquiTrap and ovitrap captures differ be-tween neighborhoods. The same pattern has beenobserved by other authors. In Belo Horizonte, Brazil,Gama et al. (2007) observed that ovitraps detected thepresence of Aedesmosquitoes during 17 wk (egg den-sity varying from 26.6% to 82.0%), whereas the Mos-quiTrap detected its presence for only 13 wk (withadult density index varying from 0 to 1.6%). Lownumber of mosquitos captured in the MosquiTrap maybe attributable to events of mosquitoes scape from thetrap, if they ßy away without landing on the sticky card(Gama et al. 2007). A study conducted in Australiawith a similar sticky trap, however, showed that stickytraps and ovitraps may detect similar infestation levels(Ritchie et al. 2003). Other potential explanations forthe relatively lower efÞciency of MosquiTrap com-pared with ovitrap, may be due to differences in at-traction. Favaro et al. (2006), though, showed that inMirassol, Sao Paulo, the sensitivity of MosquiTrapsoutdoors was 82.1%, whereas that of ovitraps was89.7%.

By comparing the three areas, we found the sub-urban and urban neighborhoods considerably moreinfested than the suburban slum. This is a surprisingresult, as the suburban slum presents all the conditionsfor mosquito development (poor housing conditions,poor water supply, and irregular garbage collection).Moreover, contrasting with the other areas, in thissuburban slum, the mosquito fauna collected in thetraps was dominated by Cx. quinquefasciatus (Table

Fig. 7. Time series of Ae. aegypti adult and egg captures. In gray, the observed data; in black, the DLM model and in red,the GAM model in three neighborhoods of Rio de Janeiro, Brazil: Higienopolis (urban), Tubiacanga (suburban), and Palmares(suburban slum).


2). This neighborhood is bordered by a polluted river,with high contents of organic matter, a preferred hab-itat for this species (Consoli and Lourenco-de-Ol-iveira 1994).

Studies carried out in different parts of the worldhave found evidence of seasonal pattern ofAe. aegyptipopulation density, by using either positivity or quan-titative trap indices (Mogi et al. 1988, Micieli andCampos 2003, Vezzani et al. 2004, Stein et al. 2005,Gama et al. 2007, Facchinelli et al. 2007, Vezzani andCarbajo 2008). The strength of the seasonal signal,however, changes with the latitude. In subtropicalregions, mosquito abundance tends to be greater dur-ing the hot months and very low (or almost absent)during the winter as observed in Buenos Aires, Ar-gentina (Vezzani et al. 2004, Vezzani and Carbajo2008). In the equatorial region, however, mosquitoseasonal variation is more subtle or inexistent, as in theAmazonian city of Manaus (03� 07� S, 59� 57� W)(Rõos-Velasquez et al. 2007).

In Rio de Janeiro, a tropical city, we found a sig-niÞcantly positive effect of air temperature at lag 1 wk,on the amount of eggs/adult captured. This effect wasfound in all three localities for both traps. Tempera-ture at this lag should impact the preadult develop-ment of the mosquito population (Schreiber 2001).Below 22Ð24�C, the effect of temperature was approx-imately linear and negative, that is, below this thresh-old, lower temperature contributes to the decline ofthe mosquito population (Fig. 6). Above 22Ð24�C, theincremental effect of temperature is not evident. Infact, during yellow fever transmission in 1908, it wasveriÞed that Ae. aegypti development and reproduc-tion were strongly affected when temperaturedropped below 20�C in Rio de Janeiro (Lourenco-de-Oliveira 2008). Moreover, temperatures above 30�Cmay have minimal impact on Ae. aegypti in the Þeld,because this mosquito may avoid excessive daytimeheat by resting in cooler, shaded, locations indoors(Schreiber 2001). Previous studies found the maxi-mum survival rates for adultAe. aegypti in the range of20Ð30�C (Rueda et al. 1990, Tun-Lin et al. 2000).Beserra et al. (2006) found the favorable temperatureforAe. aegyptidevelopment is between 21�C and 29�C,and for greater longevity and fecundity, between 22�Cand 30�C. Maciel-de-Freitas et al. (2007b) found atendency toward longer survival ofAe. aegypti femalesduring the dry winter (18.3Ð26.5�C; 44.1 mm rainfall)than in the wet and hot summer (23.3Ð29.6�C; 1.9 mm)in the suburban area here studied. Our results alsosuggest a negative effect of temperature at lag 3. Thisresult is less intuitive. A possible explanation for thiseffect is that temperature also contributes to the re-duction of oviposition sites (by evaporation) in areaswhere discarded containers such as cans, tires, andbottles are important for larvae and pupae mainte-nance, and such effect may occur at longer lags, whenthe direct effect of temperature on mosquito biologybecomes less important. In addition, previous studiesshow that long periods of high temperatures anddrought may cause high mortality of Aedes eggs(Gubler 1988; Costanzo et al. 2006).

In the literature, the effect of rainfall on mosquitodynamics has been elusive (Moore et al. 1978, Louni-bos 1981, Scott et al. 2000, Honorio and Lourenco-de-Oliveira 2001, Lourenco-de-Oliveira et al. 2004). Rain-fall should be important in areas where breeding sitesare mainly produced by precipitation. In the urbanarea (Higienopolis), we observed well maintainedyards (clean and cemented) with low potential larvalhabitats outdoors (Table 1). In the suburban area(Tubiacanga), there is a high availability of permanentcontainers, such as water tanks and metal drums (notaffected by rainfall), as well as a high abundance ofcontainers potentially affected by rainfall (Maciel-de-Freitas et al. 2007b). In the suburban slum, where themain human activity is garbage recycling, there arealso many potential breeding sites outdoors. As a con-sequence, we expected to detect a signiÞcant effect ofrainfall in the suburban area and the suburban slum,but not in urban neighborhood of Higienopolis.

Our results seem to agree with this expectation. Inboth suburban neighborhood and suburban slum, theinclusion of rainfall improved the model goodness-of-Þt (Table 5), but not in the urban area. Whenlooking at the lag distribution of rainfall effect, though,we Þnd evidence of positive effect of rainfall at longerlags in all three areas, and a short term negative effect(lag 1), only in the suburban. The positive effect ofrainfall at lags 3Ð5 can be attributed to the main impactof rainfall on population dynamics, that is, productionof new oviposition sites and egg eclosion stimulationin existing ones. This lag corresponds to the periodduring which the parent generation was ovipositingthe current (captured) generation. Rainfall, at lag 1,however, is likely to affect trap performance and thebehavior of the current mosquito generation. In thesuburban neighborhood, where most houses are sur-rounded by open yards, rainfall had a negative effecton trap index at lag 1. Increasing the availability ofoviposition sites (generated by precipitation) may re-duce the efÞciency of the traps, as they have to com-pete with more alternative oviposition sites. In theurban and the suburban slum, however, where openspaces are rarer, this effect is not observed. This resultcontrasts with those of Costa-Ribeiro et al. (2006) whosuggested that the higher availability of larval habitatsduring the rainy season would restrain dispersal ofAe.aegypti gravid females seeking for sites for ovipositionin Higienopolis and Vargem Grande (Palmares neigh-borhood), leading to low genetic diversity.

In conclusion, the main results of this study can besummarized as follows: qualitative and quantitativeindices using ovitraps or MosquiTrap detected similarsummer peaks of mosquito population dynamics, sug-gesting that both could be used for surveillance; how-ever, the ovitrap provided a more sensitive index thanMosquiTrap and should be preferable in areas withlow infestation (as in the suburban slum). Moreover,the quantitative indices were more sensitive than thequalitative indices when used for comparisons amongareas. Rainfall showed a positive effect at long lags(3Ð5 wk), potentially due to the production of newbreeding sites. At last, temperature above 22Ð24�C was


strongly associated with high mosquito density andconsequently, an increased risk of dengue transmis-sion. These results are a contribution for the devel-opment of a future climate based control and surveil-lance program for dengue fever in a hyperendemicarea.

Acknowledgments

Special thanks go to Celio da Silva Pinel, Jefferson Silva,Marcelo Vicente, Alexsandro Camargo, Celma Marinho, Lu-ciene Silva, Crissie Ferraz, Marcelo Celestino, Mauro Men-ezes, Renato Carvalho, Reginaldo Rego, Priscilla Pereira,Denise Pio, Camila Pinto, Debora Elaine, Andre Luiz, and thetechnicians of Secretaria Municipal de Saude do Rio deJaneiro for logistical support in the Þeld and laboratory. Wealso thank the members of the SAUDAVEL Project, MarõliaSa Carvalho, Aline Nobre, Paulo Chagastelles Sabroza,Steven Juliano, and Phil Lounibos for fruitful discussionsduring the elaboration of this work and Oswaldo Cruz forhelping with R programming. We thank to the people ofTubiacanga, Higienopolis, and Palmares for collaboration.This work was supported by FIOCRUZ-Fundacao OswaldoCruz, FAPERJ-Fundacao de Amparo a Pesquisa do EstadodoRiode Janeiro,CAPES-CoordenacaodeAperfeicoamentode Pessoal de Nõvel Superior, CNPq-Conselho Nacional deDesenvolvimento CientõÞco e Tecnologico.

References Cited

Akaike, H. 1974. A new look at the statistical model iden-tiÞcation. IEEE Trans. Automatic Control 19: 716Ð723.

Bartley, L.M., C. A. Donnelly, andG. P. Garnett. 2002. Theseasonal pattern of dengue in endemic areas: mathemat-ical models of mechanisms. Trans. R. Soc. Trop. Med.Hyg. 96: 387Ð397.

Beserra, E. B., Jr., F. P. Castro, J. W. Santos, T. S. Santos, andC.R.M. Fernandes. 2006. Biologia e exigencias termicasde Aedes aegypti (L.) (Diptera: Culicidae) provenientesde quatro regioes bioclimaticas da Paraõba. Neotrop. En-tomol. 35: 853Ð860.

Braga, I. A., A. C. Gomes, M. Nelson, R.C.G. Mello, D. P.Bergamaschi, and J.M.P. Souza. 2000. Comparacao en-tre pesquisa larvaria e armadilha de oviposicao para de-teccao de Aedes aegypti. Rev. Bras. Med. Trop. 33: 347Ð353.

Canyon, D. V., J.L.K. Hii, and R. Muller. 1999. Adaptationof Aedes aegypti (Diptera: Culicidae) oviposition behav-iour in response to humidity and diet. J. Insect Physiol. 45:959Ð964.

Connor, M. E., andW. M. Monroe. 1923. Stegomyia indicesand their value in yellow fever control. Am. J. Trop. Med.Hyg. 3: 9Ð19.

Consoli, R.A.G.B., andR. Lourenco-de-Oliveira. 1994. Prin-cipais mosquitos de importancia sanitaria no Brasil. Edi-tora Fiocruz, Rio de Janeiro, Brazil.

Constanzo, K. S., B. Kesavaraju, and S. A. Juliano. 2006.Condition-speciÞc competition in container mosquitoes:the role of noncompeting life-history stages. Ecology 86:3289Ð3295.

Costa-Ribeiro, M. C., R. Lourenco-de-Oliveira, and A. B.Failloux. 2006. Geographic and temporal genetics pat-terns of Aedes aegypti populations in Rio de Janeiro,Brazil. Trop. Med. Int. Health 11: 1276Ð1285.

Crans,W. J.,D. A. Sprenger, F.Mahmood. 1996. The blood-feeding habitats of Aedes sollicitans (Walker) in relationto Eastern Equine Encephalitis virus in coastal areas of

New Jersey, II. Results of experiments with caged mos-quitoes and the effects of temperature and physiologicalage on host selection. J. Vector Ecol. 21: 1Ð5.

Day, J. F., A. M. Ramsey, and J. Zhang. 1990. Environmen-tally mediated seasonal variation in mosquito body size.Environ. Entomol. 19: 469Ð473.

Eiras, A. E. 2002. Armadilha para captura de mosquitos.Patente: Privilegio e Inovacao. N. PI0203907-9, “Ar-madilha para captura de mosquitos”. 05 de set de 2002(Deposito).

Facchinelli,L.,L.Valerio,M.Pombi,P.Reiter,C.Costantini,and D. Torre. 2007. Development of a novel sticky trapfor container-breeding mosquitoes and evaluation of itssampling properties to monitor urban populations ofAedes albopictus. Med. Vet. Entomol. 21: 183Ð195.

Favaro, E. A., A. Mondini, M. R. Dibo, A.A.C. Barbosa, A. E.Eiras, and F. Chiaravalloti-Neto. 2008. Assessment ofentomological indicators ofAedes aegypti (L.) from adultand egg collections in Sao Paulo, Brazil. J. Vector Ecol. 33:8Ð16.

Favaro, E. A., M. R. Dibo, A. Mondini, A. C. Ferreira, A.A.C.Barbosa, A. E. Eiras, E.A.M.F. Barata, and F. Chiaraval-loti-Neto. 2006. Physiological state ofAedes (Stegomyia)aegypti mosquitoes captured with MosquiTrapsTM inMirassol, Sao Paulo, Brazil. J. Vector Ecol. 31: 285Ð291.

Fay, R.W., andD. A. Eliason. 1966. A preferred ovipositionsite as surveillance method forAedes aegypti.Mosq. News26: 531Ð535.

Focks, D. A. 2003. A review of entomological samplingmethods and indicators for dengue vectors. In Tropicaldisease research. World Health Organization, Geneva,Switzerland.

Focks, D. A., D. C.Haile, E. Daniels, andG. A.Moun. 1993a.Dynamics life table model for Aedes aegypti: analysis ofthe literature and model development. J. Med. Entomol.30: 1003Ð1018.

Focks,D. A.,D. C.Haile, E.Daniels, andG. A.Moun. 1993b.Dynamics life table model for Aedes aegypti: simulationresults. J. Med. Entomol. 30: 1019Ð1029.

Gama, R. A., E. M. Silva, I. M. Silva, M. C. Resende, and A. E.Eiras. 2007. Evaluation of the sticky MosquiTRAPTM fordetecting Aedes (Stegomyia) aegypti (L.) (Diptera: Cu-licidae) during the dry season in Belo Horizonte, MinasGerais, Brazil. Neotrop. Entomol. 36: 294Ð302.

Gubler,D. J. 1988. Dengue, pp. 223Ð260. InT. P. Monath TP[ed.], The arboviruses: epidemiology and ecology, vol-ume II. CRC, Boca Raton, FL.

Halstead, S. B. 2008. Dengue virusÐmosquito interactions.Annu. Rev. Entomol. 53: 273Ð291.

Hawley, W. A. 1985. The effect of larval density on adultlongevity of a mosquito, Aedes sierrensis: epidemiologicalconsequences. J. Anim. Ecol. 54: 955Ð964.

Honorio, N. A., and R. Lourenco-de-Oliveira. 2001. Fre-quencia de larvas e pupas de Aedes aegypti e Aedes al-bopictus em armadilhas, Brasil. Rev. Saude Publ. 34: 385Ð391.

Honorio,N.A., P.H.Cabello,C.T.Codeco, andR.Lourenco-de-Oliveira. 2006. Preliminary data on the performanceof Aedes aegypti and Aedes albopictus immatures devel-oping in water-Þlled tires in Rio de Janeiro. Mem. Inst.Oswaldo Cruz 101: 225Ð228.

Honorio, N. A., W. C. Silva, P. J. Leite, J. M. Goncalves, L. P.Lounibos, and R. Lourenco-de-Oliveira. 2003. Dispersalof Aedes aegypti and Aedes albopictus (Diptera: Culici-dae) in an urban endemic dengue area in the state of Riode Janeiro, Brazil. Mem. Inst. Oswaldo Cruz 98: 191Ð198.

[ IBGE] InstitutoBrasileiro deGeografia eEstatıstica. 2000.Censo demograÞco 2000 agregados por setores censi-


tarios dos resultados do universo, 11Ð417E. InstitutoBrasileirodeGeograÞaeEstatõstica,Riode Janeiro,Brasil.

Lounibos, L. P. 1981. Habitat segregation among Africantreehole mosquitoes. Ecol. Entomol. 6: 129Ð154.

Lourenco-de-Oliveira, R. 2008. Rio de Janeiro againstAedesaegypti: yellow fever in 1908 and dengue in 2008Ðedito-rial. Mem. Inst. Oswaldo Cruz 103: 627Ð628.

Lourenco-de-Oliveira,R.,M.G.Castro,M.A.Braks, andL.P.Lounibos. 2004. The invasion of urban forest by denguevectors in Rio de Janeiro. J. Vector Ecol. 29: 94Ð100.

Luz, C., M.H.H. Tai, A. H. Santos, and H.H.G. Silva. 2008.Impact of moisture on survival of Aedes aegypti eggs andovicidal activity of Metarhizium anisopliae under labora-tory conditions. Mem. Inst. Oswaldo Cruz 103: 214Ð215.

Luz, P. M., C. T. Codeco, E. Massad, and C. J. Struchiner.2003. Uncertainties regarding dengue modelling in Riode Janeiro, Brazil. Mem. Inst. Oswaldo Cruz 98: 871Ð878.

Maciel-de-Freitas,R,W.A.Marques,R.C.Peres, S. P.Cunha,and R. Lourenco-de-Oliveira. 2007b. Variation in Aedesaegypti (Diptera: Culicidae) container productivity in aslum and a suburban district of Rio de Janeiro during dryand wet seasons. Mem. Inst. Oswaldo Cruz 102: 489Ð496.

Maciel-de-Freitas, R., C. T. Codeco, and R. Lourenco-de-Oliveira. 2007a. Daily survival rates and dispersal ofAedes aegypti females in Rio de Janeiro, Brazil. Am. J.Trop. Med. Hyg. 76: 659Ð665.

Melo, P.R.S., E.A.G. Reis, I. A. Ciuffo,M.Goes, R. E. Blanton,and M. G. Reis. 2007. The dynamics of dengue virusserotype 3 introduction and dispersion in the state ofBahia, Brazil. Mem. Inst. Oswaldo Cruz 102: 905Ð912.

Micieli, M. V., and R. E. Campos. 2003. Oviposition activityand seasonal pattern of a population ofAedes (Stegomyia)aegypti (L.) (Diptera: Culicidae) in subtropical Argen-tina. Mem. Inst. Oswaldo Cruz 98: 659Ð663.

Ministerio da Saude. 2008. Servico de Vigilancia em Saude-SVS. (http://portal.saude.gov.br/portal/saude/area).

Mogi,M., C. Khamboonruang,W. Choochote, and P. Suwan-panit. 1988. Ovitrap surveys of dengue vector mosqui-toes in Chiang Mai, northern Thailand: seasonal shifts inrelative abundance of Aedes albopictus and Ae. aegypti.Med. Vet. Entomol. 2: 319Ð324.

Moore,C.G., B. L.Cline, E. Ruiz-Tiben,D. Lee,H.Romney-Joseph, and E. Rivera-Correa. 1978. Aedes aegypti inPuerto Rico: environmental determinants of larval abun-dance and relation to dengue virus transmission. Am. J.Trop. Med. Hyg. 27: 1255Ð1231.

Morato, V.C.G., M. G. Teixeira, A. C. Gomes, D. P. Berga-maschi, and M. Barreto. 2005. Infestation of Aedes ae-gypti estimated by oviposition traps in Brazil. Rev. SaudePubl. 39: 553Ð558.

Nogueira,R.M.R.,M.P.Miagostovich,A.M.B.Filippis,M.A.S.Pereira, andH. G. Schatzmayr. 2001. Dengue virus type3 in Rio de Janeiro, Brazil. Mem. Inst. Oswaldo Cruz 96:925Ð926.

Nogueira, R.M.R.,M. P.Miagostovich, andH.G. Schatzmayr.2002. Dengue virus in Brazil. Dengue Bull. 26: 1Ð10.

Osanai, C. H., A.P.A. Travassos-da-Rosa, A. T. Tang, R. S.Amaral, A.D.C. Passos, and P. I. Tauil. 1983. Surto dedengue em Boa Vista, Roraima. Rev. Inst. Med. Trop. SaoPaulo 25: 53Ð54.

Pant, C. P., and M. Yasuno. 1973. Field studies on thegonotrophic cycle ofAedes aegypti in Bangkok, Thailand.J. Med. Entomol. 10: 219Ð223.

Parker, M. B. 1986. Hatchability of eggs of Aedes taenio-rhynchus (Diptera: Culicidae): effects of different tem-peratures and photoperiods during embryogenesis. Ann.Entomol. Soc. Am. 79: 925Ð930.

RDevelopment Core Team. 2006. R: a language and environ-ment for statistical computing. (http://www.R-project.org).

Regis, L., A.M.Monteiro,M.A.V.Melo-Santos, J. C. Silveira, Jr.,A. F. Furtado,R.V.Acioli,G.M. Santos,M.Nakazawa,M. S.Carvalho, P. J. Ribeiro, Jr., andW. V. Souza. 2008. Devel-oping new approaches for detecting and preventing Aedesaegypti population outbreaks: basis for surveillance, alertand control system. Mem. Inst. Oswaldo Cruz 103: 50Ð59.

Reiter, P.,M.A. Amador, andN.Colon. 1991. Enhancementof the CDC ovitrap with hay infusions for daily monitor-ing of Aedes aegypti populations. J. Am. Mosq. ControlAssoc. 7: 52Ð55.

Rıos-Velasquez, C. M., C. T. Codeco, N. A. Honorio, P. C.Sabroza, M. Moresco, I.C.L. Cunha, A. Levino, L. M.Toledo, and S. B. Luz. 2007. Distributions of denguevectors in neighborhoods with different urbanizationtypes of Manaus, state of Amazonas, Brasil. Mem. Inst.Oswaldo Cruz 102: 617Ð623.

Ritchie, S. A., S. Long, A. Hard, C. E. Webb, and R. C. Russel.2003. An aduticidal sticky ovitrap for sampling container-breeding mosquitoes. J. Am. Mosq. Control Assoc. 19: 235Ð242.

Ritchie, S. A., L. Sharron, G. Smith, A. Pyke, and T. Knox.2004. Entomological investigations in a focus of denguetransmission in Cairns, Queensland, Australia, by usingthe sticky ovitraps. J. Med. Entomol. 41: 1Ð4.

Rueda, L.M., K. J. Patel, R. C. Axtell, andR. E. Stinner. 1990.Temperature-dependent development and survival ratesof Culex quinquefasciatus and Aedes aegypti (Diptera:Culicidae). J. Med. Entomol. 27: 892Ð898.

Schatzmayr, H. G. 2000. Dengue situation in Brazil by year2000. Mem. Inst. Oswaldo Cruz 95: 179Ð181.

Schreiber, K. V. 2001. An investigation of relationships be-tween climate and dengue using a water budgeting tech-nique. Int. J. Biometeorol. 45: 81Ð89.

Scott, T. W., A. C. Morrison, L. H. Lorenz, G. C. Clark, D.Strickman, P. Kittayapong,H. Zhou, and J. Edman. 2000.Longitudinal studies of Aedes aegypti (Diptera: Culici-dae) in Thailand and Puerto Rico: population dynamics.J. Med. Entomol. 37: 77Ð88.

[SESDEC/RJ] Secretaria de Defesa Civil do Rio de Ja-neiro. 2008. Acoes em Saude. Relatorio de casos dedengue. (http://www.saude.rj.gov.br/Docs/Acoes/dengue/Relatorio.htm).

[SMS/RJ] Secretaria Municipal de Saude do Rio de Janeiro.2008. Casos de incidencia de Dengue por bairro e mes, epor bairro e semana no Municõpio do Rio de Janeiro.(http://www.saude.rio.rj.gov.br).

Souza, I.C.A., R.P.T. Vianna, and R. M. Moraes. 2007. Mod-elagem da incidencia do dengue na Paraõba, Brasil, pormodelos de defasagem distribuõda. Cad. Saude Publ. 23:2623Ð2630.

Stein, M., G. I. Oria, W. R. Almiron, and J. Willener. 2005.Fluctuacion estacional de Aedes aegypti en Chaco, Ar-gentina. Rev. Saude Publ. 39: 559Ð564.

Tauil, P. 2001. Urbanizacao e ecologia do dengue. Cad.Saude Publ. 17: 99Ð102.

Teixeira, M. G., M.C.N. Costa, G. Coelho, andM. L. Barreto.2008. Recent shift in age pattern of dengue hemorrhagicfever, Brazil. Emerg. Infect. Dis. 14: 1663.

Tun-Lin, W., T. R. Burkot, and B. H. Kay. 2000. Effects oftemperature and larval diet on development rates andsurvival of the dengue vector Aedes aegypti in northQueensland, Australia. Med. Vet. Entomol. 14: 31Ð37.

Vezzani, D., and A. E. Carbajo. 2008. Aedes aegypti, Aedes al-bopictus, and dengue in Argentina: current knowledge andfuture directions. Mem. Inst. Oswaldo Cruz 103: 66Ð74.


Vezzani, D., S. M. Velasquez, and N. Schweigmann. 2004.Seasonal pattern of abundance ofAedes aegypti (Diptera:Culicidae) in Buenos Aires city, Argentina. Mem InstOswaldo Cruz 99: 351Ð356.

Wearing, H. J., and P. Rohani. 2006. Ecological and immu-nological determinants of dengue epidemics. Epub 103:11802Ð11807.

Wood, S. N. 2006. Generalized additive models: an intro-duction with Chapman & Hall/CRC, Boca Raton, FL.

Wu, P.-C., H.-G. Guo, S.-C. Lung, C.-Y. Lin, and H.-J. Su.2007. Weather as an effective predictor for occurrence ofdengue fever in Taiwan. Acta Trop. 103: 50Ð57.

Received 21 August 2008; accepted 14 April 2009.


s ,d temporal distribution of aedes aegypti in different...

Documents