SEARCH

SEARCH BY CITATION

Keywords:

  • dengue;
  • Aedes aegypti ;
  • spatial;
  • temporal;
  • Cairns

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Objectives

To identify the meteorological drivers of dengue vector density and determine high- and low-risk transmission zones for dengue prevention and control in Cairns, Australia.

Methods

Weekly adult female Ae. aegypti data were obtained from 79 double sticky ovitraps (SOs) located in Cairns for the period September 2007–May 2012. Maximum temperature, total rainfall and average relative humidity data were obtained from the Australian Bureau of Meteorology for the study period. Time series–distributed lag nonlinear models were used to assess the relationship between meteorological variables and vector density. Spatial autocorrelation was assessed via semivariography, and ordinary kriging was undertaken to predict vector density in Cairns.

Results

Ae. aegypti density was associated with temperature and rainfall. However, these relationships differed between short (0–6 weeks) and long (0–30 weeks) lag periods. Semivariograms showed that vector distributions were spatially autocorrelated in September 2007–May 2008 and January 2009–May 2009, and vector density maps identified high transmission zones in the most populated parts of Cairns city, as well as Machans Beach.

Conclusion

Spatiotemporal patterns of Ae. aegypti in Cairns are complex, showing spatial autocorrelation and associations with temperature and rainfall. Sticky ovitraps should be placed no more than 1.2 km apart to ensure entomological coverage and efficient use of resources. Vector density maps provide evidence for the targeting of prevention and control activities. Further research is needed to explore the possibility of developing an early warning system of dengue based on meteorological and environmental factors.

Objectifs

Identifier les déterminants météorologiques de la densité du vecteur de la dengue et déterminer les zones de haut et de bas risque de transmission pour la prévention et le control de la dengue à Cairns, en Australie.

Méthodes

Les données hebdomadaires sur les femelles adultes d’Ae. aegypti ont été obtenues à partir de 79 pondoirs pièges à doubles collants situés à Cairns durant la période de septembre 2007 à mai 2012. La température maximale, les précipitations totales et la moyenne des données d'humidité relative ont été obtenues auprès du Bureau Australien de la Météorologie pour la période d’étude. Des modèles de distribution non linéaire des séries de latence ont été utilisés pour évaluer la relation entre les variables météorologiques et la densité du vecteur. L'autocorrélation spatiale évaluée par semivariography et par krigeage ordinaire a été entreprise afin de prédire la densité du vecteur à Cairns.

Résultats

La densité d’Ae. aegypti a été associée avec la température et les précipitations. Cependant, ces relations différaient entre les périodes de latence courte (0–6 semaines) et longue (0–30 semaines). Les semivariogrammes ont montré que les distributions du vecteur étaient spatialement autocorrélées dans la période de septembre 2007 à mai 2008 et de janvier 2009 à mai 2009, et les cartes de densité du vecteur ont identifié des zones de haute transmission dans les zones les plus peuplées de la ville de Cairns, ainsi que de Machans Beach.

Conclusion

Les tendances spatio-temporelles d’Ae. aegypti à Cairns sont complexes, montrant une autocorrélation spatiale et des associations avec la température et les précipitations. Les pondoirs pièges collants ne doivent pas être placés à plus d'1,2 km les uns des autres afin d'assurer une couverture entomologique et l'utilisation efficace des ressources. Les cartes de densité vectorielle fournissent des preuves pour le ciblage des activités de prévention et de contrôle. Des recherches supplémentaires sont nécessaires pour explorer la possibilité de développer un système d'alerte précoce de la dengue basée sur les facteurs météorologiques et environnementaux.

Objetivos

Identificar los factores meteorológicos con influencia sobre la densidad del vector del dengue y determinar las zonas de alto y bajo riesgo de transmisión para la prevención del dengue y su control en Cairns, Australia.

Métodos

Semanalmente se obtuvieron datos sobre hembras adultas de Ae. aegypti data de 79 trampas pegajosas dobles (TP) localizadas en Cairns entre Septiembre 2007 – Mayo 2012. Datos para el periodo del estudio de temperatura máxima, precipitación total y datos promedio de humedad relativa se obtuvieron de la Oficina Australiana de Meteorología. Se utilizaron modelos de retardos distribuidos de series temporales no lineales para evaluar la relación entre las variables meteorológicas y la densidad vectorial. La autocorrelación espacial se evaluó mediante semivariografía y se realizó un kriging ordinario para predecir la densidad vectorial en Cairns.

Resultados

La densidad de Ae. aegypti estaba asociada con la temperatura y las precipitaciones. Sin embargo, esta relación difería entre periodos de tiempo cortos (0–6 semanas) y largos (0–30 semanas). Los semivariogramas mostraban que las distribuciones vectoriales estaban espacialmente autocorrelacionadas entre Septiembre 2007 – Mayo 2008 y Enero 2009 – Mayo 2009 y los mapas de densidad vectorial identificaron áreas de transmisión alta en las partes más pobladas de la ciudad de Cairns, al igual que en la playa de Machans.

Conclusión

Los patrones espaciotemporales de Ae. aegypti en Cairns son complejos, mostrando una autocorrelación espacial y asociaciones con la temperatura y las precipitaciones. Las trampas pegajosas deberían de colocarse con un máximo de 1.2 km de separación entre ellas, con el fin de asegurar la cobertura entomológica y un uso eficiente de recursos. Los mapas de densidad vectorial aportan evidencia para afinar en las actividades de prevención y de control. Se requieren más estudios para explorar la posibilidad de desarrollar un sistema de advertencia temprana para el dengue, basado en factores meteorológicos y ambientales.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Dengue fever (DF) is an increasing public health concern, with a 30-fold increase in global incidence in the past 50 years (World Health Organization 2012a). Each year there are approximately 50–100 million reported cases of DF and more than 20 000 reported deaths from dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS) (World Health Organization 2012b). In Australia, epidemics occur in northern Queensland, with most cases in Cairns (Williams et al. 2010). Epidemics result from viraemic travellers importing DF during the wet season, when environmental conditions are favourable for breeding and survival of the DF vector, Aedes aegypti. Molecular epidemiology confirms a pattern of repeated imports seeding fresh outbreaks (Ritchie et al. 2002). Small outbreaks of locally acquired DF take place almost yearly in northern Queensland, with large outbreaks occurring every 4–5 years. Locally acquired DF cases in northern Queensland increased from 27 in 1991 to 69 in 2011, with the highest number of cases recorded in 2008–2009 (n = 1037) (Scott Ritchie, unpublished observation).

In the absence of a publicly available vaccine, and despite questionable evidence for the association between vector density and virus transmission (Scott et al. 2003), entomological surveillance is considered a key strategy for the control of DF (Ooi et al. 2006; Canyon 2007; Morrison et al. 2008). In Cairns, entomological surveillance is undertaken weekly and includes monitoring double sticky ovitraps (SOs), which are designed to attract gravid adult female Aeaegypti mosquitoes (Ritchie et al. 2004; Chadee & Ritchie 2010).

When the number of mosquitoes collected in SOs exceeds one adult female Ae. aegypti per day or health authorities are notified of a DF case, control measures are implemented: physical removal of breeding sites, installation of lethal ovitraps, interior residual spraying (IRS) and community awareness and education campaigns. Testing of Wolbachia bacteria-infected mosquitoes, which decreases vector competence, is underway in Cairns; however, it may be a number of years before Ae. aegypti populations are entirely infected with Wolbachia (Hoffmann et al. 2011). In the meantime, targeting of vector control activities may be improved by the development of an early warning system using prospective meteorological data to predict areas of high vector density, using a proactive (preventive) rather than a reactive (control) framework (Racloz et al. 2012).

Evidence for the impact of meteorological factors on vector density is inconclusive (Jansen & Beebe 2010). An early study found that rainfall and humidity, but not temperature, were associated with increases in Ae. aegypti larval populations in India (Biswas et al. 1993). In a 2006 study, humidity and temperature were positively correlated with entomological indices (Favier et al. 2006), while more recent studies showed rainfall to be the only meteorological factor investigated that was a significant driver of vector density (Dibo et al. 2008; Fávaro et al. 2008; Miyazaki et al. 2009; Wan et al. 2009). Using lag periods to determine the influence of previous meteorological conditions on vector density reveals additional complexities. For example, rainfall, temperature and humidity were all significantly associated with larval density at different time points from 0–4 months prior to entomological collection date in Taiwan (Wu et al. 2007; Chen et al. 2010). Knowledge of temporal lags is particularly important for the design of climate-based surveillance systems, where identification of suitable climatic conditions at a given time point can trigger interventions to prevent subsequent vector-borne disease outbreaks.

Recent advances in geographical information systems (GISs) and geostatistical techniques have enhanced our understanding of the spatial dynamics of Aedes populations (Eisen & Lozano-Fuentes 2009; Higa 2011). Most commonly, GISs have been used for visualising the distribution and density of vector breeding sites in a given location (Sithiprasasna et al. 2004; Moreno-Sanchez et al. 2006; Tsuda et al. 2006; Chang et al. 2009). More advanced techniques include identifying and predicting high-risk transmission zones (Carbajo et al. 2001; Ali et al. 2003; Getis et al. 2003; Chansang & Kittayapong 2007).

This study investigates temporal relationships between rainfall, temperature, humidity and Ae. aegypti to determine the meteorological drivers of DF transmission in Cairns. It also examines spatial patterns of Ae. aegypti and uses interpolation techniques to predict vector density in locations without SOs to help identify high-risk transmission zones. Study implications are discussed in the context of integrated dengue surveillance and sustainable prevention and control in Cairns.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Data collection

Weekly Ae. aegypti surveillance data (numbers of adult female Ae. aegypti mosquitoes captured) were obtained from 79 double SOs located in Cairns for the period September 2007–May 2012 (Figure 1). Cairns is a low-lying coastal area with seasonal high temperatures and rainfall from November to May and cooler, drier conditions from June to October.

image

Figure 1. Geographical distribution of 79 sticky ovitraps in Cairns, Australia.

Download figure to PowerPoint

Each SO comprises two black buckets. The top bucket is lined with a polybutylene adhesive panel; the base bucket is filled with water and contains an organic attractant (e.g. lucerne) and a pellet of the insect growth regulator methoprene, to prevent mosquito production in SOs (Ritchie et al. 2003). Dry glue was used in the SOs from May 2008–December 2008 and was later found to be less efficacious than the polybutylene; therefore, this period was excluded from all analyses. Data on numbers of male Ae. aegypti were also excluded from analyses because they do not transmit DF.

Meteorological data were obtained from the Australian Bureau of Meteorology (BoM) for Cairns Aero (temperature and rainfall data) and Cairns Severin Street (humidity data) weather stations for the study period. Daily observations for maximum temperature and total rainfall were retrieved from the BoM website (http://www.bom.gov.au/climate/data/), and average daily relative humidity data were accessed via a data request. Daily observations were averaged by week (Figure 2).

image

Figure 2. Weekly collected Ae. aegypti, average relative humidity (%), maximum temperature (°C) and total rainfall (mL), September 2007–May 2012. Missing data in SOs is due to the use of dry glue for the period May 2008–December 2008.

Download figure to PowerPoint

Spatial and temporal analyses of data were undertaken separately to examine patterns and relationships in the data more comprehensively.

Temporal statistical analysis

Preliminary associations between the weekly number of female adult Ae. aegypti, total rainfall, maximum temperature and mean relative humidity were determined by cross-correlation plots using a moving average filter to decompose time trends. Initially, a generalised linear model (GLM) assuming a Poisson-distributed outcome (mosquito counts) with a log link function was developed to inform the development of a distributed lag nonlinear model (DLNM). Because the total number of female adult Ae. aegypti was overdispersed relative to the Poisson distribution, a DLNM with a negative binomial distribution and a log link was adopted. A natural cubic spline DLNM was then developed to examine lagged effects of meteorological factors on Ae. aegypti density. The Akaike information criterion (AIC) was used to select the most parsimonious model. The preliminary GLM analysis justified the use of a negative binomial model for the DLNM, which was of the form:

  • display math
  • display math

where Yt is the observed number of adult female Ae. aegypti at week t; α is the intercept; t is the week of the observation; Tt is a matrix obtained by applying the DLNM to maximum temperature; Rt is a matrix obtained by applying the DLNM to total rainfall; and RHt is a matrix obtained by applying the DLNM to mean relative humidity; l is the lag weeks; and β1–β5 are the vector of coefficients for Tt,l, Rt,l, RHt,l, sin and cos, respectively. S(.) is a natural cubic spline and time is the time trend smoothing function.

Models were fitted for a polynomial of degrees 2–5 to determine the overall effects of temperature, rainfall and humidity; a degree two polynomial (quadratic term) produced the lowest AIC score and was used in the final model. Variables were centred on the mean value, and a natural cubic spline smoothing function was used to correct for seasonality. The time trend was controlled by smoothing calendar time (4 years) with 5 degrees of freedom, determined by model comparison for 5, 6 and 7 degrees of freedom. The terms sin(2πt/52) and cos(2πt/52) were included in the model to control for the annual periodicity of Ae. aegypti density (Stolwijk et al. 1999).

Lag periods were categorised as short term (0–6 weeks) and long term (0–30 weeks). Short-term lag periods were included in the model to account for the life cycle of Ae. aegypti, which generally takes 6 weeks or less. Long-term lags were included in the model to explore the influence of meteorological variables on Ae. aegypti populations over time and validate findings from a recent study that used data from 11 BG-sentinel traps to determine the feasibility of temporal predictive modelling using meteorological indices for dengue transmission in Cairns (Azil et al. 2010).

Statistical analyses were undertaken using Stata/SE11 and the dlnm package in R statistical software, version 2.15.1 (R Foundation for Statistical Computing, Vienna Austria).

Spatial statistical analysis

Spatial autocorrelation, a primary assumption of interpolation, was assessed by generating exponential binned semivariograms using the variog function in the GeoR package in R for the periods: September 2007–May 2008; January 2009–May 2009; June 2009–May 2010; June 2010–May 2011; and June 2011–May 2012. Data comprised the mean weekly number of adult female Ae. aegypti at each of the 79 georeferenced SO locations across each time period. Covariance parameters were estimated by fitting weighted least squares (WLS) models to the semivariograms using the variofit function in the GeoR package. Resulting summary statistics were used to identify the precise distance of spatial autocorrelation in the data sets (note, one decimal degree is equivalent to approximately 111 km at the equator).

Kriging is a set of spatial interpolation techniques that create a prediction surface based on measured values and distance to prediction locations, facilitated by a spatial autocorrelation model using semivariogram covariance parameters (Pfeiffer et al. 2008). Ordinary kriging was selected as the preferred model because the unknown mean is estimated for each location based on nearby data, providing flexibility for prediction. Conversely, simple kriging assumes a known mean for all locations and is defined prior to modelling.

A rectangular grid of 4900 prediction locations (70 × 70), each measuring 120 m by 180 m, covering the study area was created to predict Ae. aegypti numbers at locations without SOs. Ordinary kriging assuming a Poisson normal model was performed on the grid using the pois.krige and pois.glm.control functions in the GeoRglm package in R, and Aedes density estimates (counts) were obtained for each prediction location. A Monte Carlo Markov Chain (MCMC) Metropolis–Hastings algorithm based on Langevin–Hastings updates and parameterisation specified in semivariography was employed to simulate the conditional distributions for gridded locations given data from SOs (Papaspiliopoulus et al. 2003). The number of MCMC iterations performed was 1000, thinning was specified as 1, burn-in set to 0, and, to ensure approximately 60 per cent of the proposals were accepted as required by Langevin–Hastings updates, the proposal variance was 0.6 (Christensen & Waagepetersen 2002). Further details of ordinary kriging using R are available elsewhere (Christensen & Ribeiro 2012). The gridded kriging estimates were exported to ArcMap via raster image files created in the rgdal package in R.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Temporal analysis of meteorological factors

Patterns of Ae. aegypti in Cairns varied from 2007 to 2012 (Figure 2). Weekly Ae. aegypti numbers peaked in 2007–2008 and decreased over the study period. Trends in temperature reflected trends in Ae. aegypti and were confirmed by the cross-correlation moving averages graph (Figure 3). Rainfall and humidity demonstrated moderate associations with Ae. aegypti. Temperature was the only significant predictor of Ae. aegypti in the negative binomial model (RR = 1.20, = 6.64e-14).

image

Figure 3. Cross correlation moving averages: Ae. aegypti and total rainfall, maximum temperature and mean relative humidity.

Download figure to PowerPoint

In the final DLNM model, rainfall at two lag degrees of freedom demonstrated a significant inverse relationship with Ae. aegypti. More simply, increased rainfall resulted in significantly fewer Ae. aegypti (RR = 0.52, = 0.002). Findings for temperature (RR = 1.31, = 0.054) and humidity (RR = 0.92, = 0.56) were not significant. However, visual examination of trends at differing lag periods revealed a more complex association between the meteorological factors and vector density (Figure 4).

image

Figure 4. Relative risks by short term (0–6 weeks) and long term (0–30 weeks) lag periods for total rainfall, maximum temperature and mean relative humidity, using a natural cubic spline distributed lag non-linear model.

Download figure to PowerPoint

Patterns of rainfall over short-term lags (0–6 weeks) were fairly stable. However, long-term trends showed increased rainfall from 0 to 15 weeks lag resulted in a significantly higher number of Ae. aegypti, and increased rainfall from 16 to 30 weeks resulted in significantly lower number of Ae. aegypti. Temperature from 0 to 6 weeks lag showed a decreasing – but not significant – association. Over long-term lags, increased temperature was significantly associated with increased vector density and from 25 to 30 weeks. For humidity, short-term lags demonstrated an increasing association, which was statistically significant from 3 to 5 weeks. Long-term lags also showed an increasing association and were significant from 15 to 25 weeks.

Spatial analysis

Spatial autocorrelation was only apparent in the vector density data sets from September 2007 to May 2008 and from January 2009 to May 2009 (Figure 5), and the spatial analysis only considered these time periods. Semivariograms demonstrated spatial autocorrelation up to a distance of 0.0118 decimal degrees for September 2007–May 2008 and 0.0173 decimal degrees for January 2009–May 2009. More simply, the number of Ae. aegypti at a given SO location was autocorrelated with the number of mosquitoes at other SOs located within 1.2 km and 1.7 km for September 2007–May 2008 and January 2009–May 2009, respectively. This spatial autocorrelation enables prediction of Ae. aegypti density to locations without SOs via interpolation (kriging).

image

Figure 5. Semivariogram demonstrating spatial autocorrelation up to lag distance of 0.0118 decimal degrees (equivalent to approximately 1.2 km at the equator) for September 2007–May 2008, and 0.0173 decimal degrees (approximately 1.7 km) for January 2009–May 2009.

Download figure to PowerPoint

Figures 6 and 7 present interpolated densities of Ae. aegypti in Cairns and Machans Beach for September 2007–May 2008 and January 2009–May 2009. Areas of high vector density were similar between years and were mostly confined to the Cairns city area, with an additional high density zone in the less-populated Machans Beach area.

image

Figure 6. Map of predicted Ae. aegypti abundance for Cairns using ordinary kriging, September 2007–May 2008 and January 2009–May 2009.

Download figure to PowerPoint

image

Figure 7. Map of predicted Ae. aegypti abundance for Machans Beach using ordinary kriging, September 2007–May 2008 and January 2009–May 2009.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

These results build on findings from a recent temporal study undertaken in Cairns, to present a comprehensive assessment of DF transmission risk (Azil et al. 2010). Azil et al. (2010) found that increasing temperature 6 months prior to vector collection resulted in higher adult female Ae. aegypti density, suggesting that egg diapause and overwinter survival occur in Ae. aegypti populations in Cairns. However, the study was limited to 11 BG-sentinel traps and did not include spatial analysis. Our results confirmed that an increase in temperature 25–30 weeks prior to vector collection date resulted in a statistically significant increase in the number of Ae. aegypti found in SOs (Figure 4).

We also identified a significant inverse relationship between rainfall and Ae. aegypti density, which may be due to heavy rainfall increasing the number of water-filled containers suitable for oviposition and thus reducing the number of gravid female mosquitoes selecting a SO as a suitable breeding container. Heavy rainfall causing flushing of breeding sites may also contribute to reduced vector density; however, evidence suggests that larval and pupal Ae. aegypti populations are only slightly affected by heavy rainfall, perhaps due to adaptation to tropical habitats over time (Koenraadt & Harrington 2008). Previous studies found rainfall was a contributing factor in the transmission of DF; however, these studies focused on DF cases and did not include data on entomological indices (Hurtado-Diaz et al. 2007; Banu et al. 2011; Hu et al. 2011, 2012). Meteorological findings may inform DF management in Cairns, where control activities (e.g. physical removal of breeding sites) are scaled up following periods low rainfall and high temperature.

We used robust spatial interpolation methods to predict the density of Ae. aegypti in Cairns for improved targeting of prevention and control activities. To our knowledge, this is the first study to produce vector density (risk) maps in Cairns. As expected, risk maps showed high vector density zones in Cairns city area, where human population density is greatest: consistent with findings from a recent study in Colombia demonstrating the importance of human population density in DF transmission (Padmanabha et al. 2012). A high-risk zone seen in the less-populated Machans Beach area may be due to a profusion of vector breeding sites. The risk zone was slightly smaller and less dense in January-May 2009 compared to September 2007-May 2008, possibly demonstrating the impact of control efforts in this area following the 2008–2009 epidemic.

The consistency of risk zones between maps indicates that these areas are probably persistent breeding sites and should be targeted by the Queensland Dengue Action Response Team (DART) for regular control activities including the identification and elimination of breeding sites. Moreover, DART can direct resources away from low-risk transmission zones to these areas when required. Other studies have attempted to predict risk zones for Ae. aegypti in Australia; however, they focused on future risk for the whole country and did not include small-scale predictions for DF management in Cairns (Kearney et al. 2009; Williams et al. 2010).

There was no observed spatial autocorrelation in data sets for periods from late 2009 to 2012, which may have been due to decreases in Ae. aegypti numbers and a disruption of the spatial dynamics of vector populations following the expansion of control activities in 2009. In September 2007–May 2008, spatial autocorrelation was found up to a distance of 1.2 km, and in January 2009–May 2009, spatial autocorrelation was identified up to a distance of 1.7 km. These results indicate that SOs should be placed no further than 1.2 km apart to ensure that all of the spatial variations in vector density are captured by the sampling method. Denser sampling could be undertaken in areas of specific interest to capture finer-scale heterogeneity and facilitate targeted control. This information is useful for DART, who previously distributed SOs throughout Cairns according to the location of prior DF epidemic zones and human population density (Williams et al. 2006). Additionally, this information may be used to facilitate the distribution of SOs for DF vector surveillance in other parts of Queensland (e.g. Townsville), if required.

Our results showed the number of Ae. aegypti decreasing over the study period (2007–2012). Yet, unless prevention and control programmes are improved, these populations will resurge. Evidence supporting the correlation between vector density and DF infection is weak, with experts suggesting that even when Ae. aegypti population density is low, DF transmission can occur due to Ae. aegypti's regular blood meal feeding, which increases the likelihood of transmitting the disease (Kuno 1997). However, without other public health prevention tools such as vaccination, improving DF surveillance, prevention and control programmes with the aim of decreasing vector density should be prioritised (Scott et al. 2003). Indeed, the geographic expansion of Ae. albopictus from the Torres Strait to other parts of Australia, particularly Queensland, in the near future may result in higher vector density and increased DF transmission risk (Russell et al. 2005).

Spatial decision support systems (SDSSs) are increasingly recognised as essential for managing vector-borne diseases (Duncombe et al. 2012; Kelly et al. 2012). SDSSs utilise a range of routinely collected data and expert knowledge to explore spatiotemporal patterns of disease, including the prediction of potential epidemic locations. In 2008, a SDSS was developed to inform DF control in Mexico (Lozano-Fuentes et al. 2008). A more recent study in Cairns demonstrated space–time clustering of DF transmission during an epidemic to identify key tools for use in an integrated DF management system, such as a SDSS (Vazquez-Prokopec et al. 2010). DART currently uses GIS to identify locations of SOs, map vector control responses to DF cases and plan the distribution of IRS in Cairns; however, GIS capability is not fully utilised. The development of a SDSS for Cairns – building on a number of recent studies exploring the spatial and temporal patterns of DF, including this one – would enable data collection and reporting standardisation, targeting and coordination of control strategies and facilitation of resource allocation decisions via automated user-defined reports (Eamchan et al. 1989; Eisen & Eisen 2011).

This study has some limitations. Information on wind direction and strength, shading and the availability of and competition for food resources for Ae. aegypti – important drivers of vector survival – was not included in our analyses, possibly confounding associations between meteorological factors and vector density (Tun-Lin et al. 2000). Additionally, collected Ae. aegypti mosquitoes were not serologically tested for the presence of DF virus, and DF case data and information on serotype-specific herd immunity were not included in our analyses. Without a proven association between vector density and virus transmission, it was not possible to ascertain a direct link between our vector density predictions and the risk of DF in the human population.

Despite these limitations, our findings, along with previous studies, suggest that temperature and rainfall are associated with Ae. aegypti density. Vector density maps identify potential high transmission zones for targeting of prevention and control activities, and semivariograms provide evidence for the optimal spacing of SOs in Cairns. Further research into the spatiotemporal patterns of Ae. aegypti in Cairns should involve confirmation of the association between vector density and DF risk; additional exploration of the role of meteorological factors on Ae. aegypti to progress predictive climate-based models; and the development and consequent evaluation of the feasibility of an operational SDSS to improve the efficiency of surveillance and sustainability of DF prevention and control in Cairns.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We sincerely thank the Queensland Tropical Public Health Team and Dengue Action Response Team for their assistance and advice during this study. AC is supported by an Australian National Health and Medical Research Council Career Development Award.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Ali M, Wagatsuma Y, Emch M et al. (2003) Use of a geographic information system for defining spatial risk for dengue transmission in Bangladesh: role for Aedes albopictus in an urban outbreak. American Journal of Tropical Medicine and Hygiene 69, 634640.
  • Azil AH, Long SA, Ritchie SA et al. (2010) The development of predictive tools for pre-emptive dengue vector control: a study of Aedes aegypti abundance and meteorological variables in North Queensland, Australia. Tropical Medicine & International Health 15, 11901197.
  • Banu S, Hu W, Hurst C et al. (2011) Dengue transmission in the Asia-Pacific region: impact of climate change and socio-environmental factors. Tropical Medicine and International Health 16, 598607.
  • Biswas D, Dey S, Dutta RN et al. (1993) Observations on the breeding habitats of Aedes aegypti in Calcutta following an episode of dengue haemorrhagic fever. Indian Journal of Medical Research 97, 4446.
  • Canyon D (2007) A Review of the Dengue Mosquito, Aedes aegypti (Diptera: Culicidae), in Australia. Retrieved 26 September 2012, from http://www.tropmed.org/rreh/vol1_4.htm.
  • Carbajo AE, Schweigmann N, Curto SI, Garĺn A & Bejarán (2001) Dengue transmission risk maps of Argentina. Tropical Medicine and International Health 6, 170183.
  • Chadee D & Ritchie SA (2010) Efficacy of sticky and standard ovitraps for Aedes aegypti in Trinidad, West Indies. Journal of Vector Ecology 35, 305400.
  • Chang AY, Parrales ME, Jimenez J et al. (2009) Combining Google Earth and GIS mapping technologies in a dengue surveillance system for developing countries. International journal of health geographics 8, 49.
  • Chansang C & Kittayapong P (2007) Application of mosquito sampling count and geospatial methods to improve dengue vector surveillance. American Journal of Tropical Medicine and Hygiene 77, 897902.
  • Chen S-C, Liao C-M, Chio C-P et al. (2010) Lagged temperature effect with mosquito transmission potential explains dengue variability in southern Taiwan: insights from a statistical analysis. Science of The Total Environment 408, 40694075.
  • Christensen OF & Ribeiro PJ (2012) Package ‘geoRglm’. Comprehensive R Archive Network (CRAN). Retrieved 3 October 2012, from cran.r-project.org/web/packages/geoRglm/geoRglm.pdf
  • Christensen OF & Waagepetersen R (2002) Bayesian prediction of spatial count data using generalized linear mixed models. Biometrics 58, 280286.
  • Dibo MR, Chierotti AP, Ferrari MS et al. (2008) Study of the relationship between Aedes (Stegomyia) aegypti egg and adult densities, dengue fever and climate in Mirassol, state of Sao Paulo, Brazil. Memorias do Instituto Oswaldo Cruz 103, 554560.
  • Duncombe J, Clements A, Hu W et al. (2012) Geographical information systems for dengue surveillance. The American Journal of Tropical Medicine and Hygiene 86, 753755.
  • Eamchan P, Nisalak A, Foy HM et al. (1989) Epidemiology and control of dengue virus infections in Thai villages in 1987. American Journal of Tropical Medicine and Hygiene 41, 95101.
  • Eisen L & Eisen RJ (2011) Using geographic information systems and decision support systems for the prediction, prevention, and control of vector-borne diseases. Annual Review of Entomology 56, 4161.
  • Eisen L & Lozano-Fuentes S (2009) Use of mapping and spatial and space-time modeling approaches in operational control of Aedes aegypti and dengue. PLoS Neglected Tropical Diseases 3, e411.
  • Fávaro EA, Mondini A, Dibo M et al. (2008) Assessment of entomological indicators of Aedes aegypti (L.) from adult and egg collections in São Paulo, Brazil. Journal of Vector Ecology 33, 816.
  • Favier C, Degallier N, Vilarinhos PT et al. (2006) Effects of climate and different management strategies on Aedes aegypti breeding sites: a longitudinal survey in Brasilia (DF, Brazil). Tropical Medicine and International Health 11, 11041118.
  • Getis A, Morrison AC, Gray K et al. (2003) Characteristics of the spatial pattern of the dengue vector, Aedes aegypti, in Iquitos, Peru. American Journal of Tropical Medicine and Hygiene 69, 494505.
  • Higa Y (2011) Dengue Vectors and their Spatial Distribution. Trop Med Health 39(4 Suppl), 1727.
  • Hoffmann AA, Montgomery BL, Popovici J et al. (2011) Successful establishment of Wolbachia in Aedes populations to suppress dengue transmission. Nature 476, 454457.
  • Hu W, Clements A, Williams G et al. (2011) Spatial analysis of notified dengue fever infections. Epidemiology and Infection 139, 391399.
  • Hu W, Clements A, Williams G et al. (2012) Spatial patterns and socioecological drivers of dengue fever transmission in Queensland, Australia. Environmental Health Perspectives 120, 260266.
  • Hurtado-Diaz M, Riojas-Rodriguez H, Rothenberg SJ et al. (2007) Short communication: impact of climate variability on the incidence of dengue in Mexico. Tropical Medicine and International Health 12, 13271337.
  • Jansen CC & Beebe NW (2010) The dengue vector Aedes aegypti: what comes next. Microbes and Infection 22, 22.
  • Kearney M, Porter WP, Williams C et al. (2009) Integrating biophysical models and evolutionary theory to predict climatic impacts on species' ranges: the dengue mosquito Aedes aegypti in Australia. Functional Ecology 23, 528538.
  • Kelly GC, Tanner M, Vallely A et al. (2012) Malaria elimination: moving forward with spatial decision support systems. Trends in Parasitology 28, 297304.
  • Koenraadt CJM & Harrington LC (2008) Flushing effect of rain on container-inhabiting mosquitoes Aedes aegypti and Culex pipiens (Diptera: Culicidae). Journal of Medical Entomology 45, 2835.
  • Kuno G (1997) Factors influencing the transmission of dengue viruses. In: Dengue and Dengue Hemorrhagic Fever (eds DJ Gubler & G Kuno) CAB International, New York, pp. 6188.
  • Lozano-Fuentes S, Elizondo-Quiroga D, Farfan-Ale JA et al. (2008) Use of Google Earth to strengthen public health capacity and facilitate management of vector-borne diseases in resource-poor environments. Bulletin of the World Health Organization 86, 718725.
  • Miyazaki RD, Ribeiro AL, Pignatti MG et al. (2009) Monitoring of Aedes aegypti mosquitoes (Linnaeus, 1762) (Diptera: Culicidae) by means of ovitraps at the Universidade Federal de Mato Grosso Campus, Cuiaba, State of Mato Grosso. Revista da Sociedade Brasileira de Medicina Tropical 42, 392397.
  • Moreno-Sanchez R, Hayden M, Janes C et al. (2006) A web-based multimedia spatial information system to document Aedes aegypti breeding sites and dengue fever risk along the US-Mexico border. Health Place 12, 715727.
  • Morrison AC, Zielinski-Gutierrez E, Scott TW et al. (2008) Defining challenges and proposing solutions for control of the virus vector Aedes aegypti. PLoS Med 5, e68.
  • Ooi EE, Goh KT, Gubler DJ et al. (2006) Dengue prevention and 35 years of vector control in Singapore. Emerging Infectious Diseases 12, 887893.
  • Padmanabha H, Durham D, Correa F et al. (2012) The Interactive Roles of Aedes aegypti Super-Production and Human Density in Dengue Transmission. PLoS Neglected Tropical Diseases 6, e1799 doi:10.1371/journal.pntd.0001799.
  • Papaspiliopoulus O, Roberts GO, Skold M et al. (2003) Non-Centered Parameterizations for Hierarchical Models and Data Augmentation. Oxford University Press, Oxford.
  • Pfeiffer DU, Robinson TP, Stevenson M et al. (2008) Spatial Analysis in Epidemiology, Oxford University Press, London, UK.
  • Racloz V, Ramsey R, Tong S & Hu W (2012) Surveillance of dengue fever virus: a review of epidemiological models and early warning systems. PLoS Neglected Tropical Diseases 6, e1648.
  • Ritchie SA, Hanna JN, Hills SL et al. (2002) Dengue control in North Queensland, Australia: case recognition and selective indoor residual spraying. Dengue Bulletin 26, 713.
  • Ritchie SA, Long S, Hart A et al. (2003) An adulticidal sticky ovitrap for sampling container-breeding mosquitoes. Journal of the American Mosquito Control Association 19, 235242.
  • Ritchie SA, Long S, Smith G et al. (2004) Entomological investigations in a focus of dengue transmission in Cairns, Queensland, Australia, by using the sticky ovitraps. Journal of Medical Entomology 41, 14.
  • Russell RC, Williams CR, Sutherst RW et al. (2005) Aedes (Stegomyia) albopictus–a dengue threat for southern Australia? Communicable diseases intelligence 29, 296298.
  • Scott TW, Morrison ACC, Takken W et al. (2003) Aedes aegypti density and the risk of dengue-virus transmission. Ecological Aspects for Application of Genetically Modified Mosquitoes Kluwer Academic Publishers, The Netherlands, 187209.
  • Sithiprasasna R, Patpoparn S, Attatippaholkun W et al. (2004) The geographic information system as an epidemiological tool in the surveillance of dengue virus-infected Aedes mosquitos. Southeast Asian Journal of Tropical Medicine and Public Health 35, 918926.
  • Stolwijk AM, Straatman H & Zielhuis GA et al. (1999) Studying seasonality by using sine and cosine functions in regression analysis. Journal of Epidemiology and Community Health 53, 235238.
  • Tsuda Y, Suwonkerd W, Chawprom S et al. (2006) Different spatial distribution of Aedes aegypti and Aedes albopictus along an urban-rural gradient and the relating environmental factors examined in three villages in northern Thailand. Journal of the American Mosquito Control Association 22, 222228.
  • Tun-Lin W, Burkot TR & Kay BH (2000) Effects of temperature and larval diet on development rates and survival of the dengue vector Aedes aegypti in north Queensland, Australia. Medical and Veterinary Entomology 14, 3137.
  • Vazquez-Prokopec GM, Kitron U, Montgomery B, Horne P & Ritchie SA (2010) Quantifying the spatial dimension of dengue virus epidemic spread within a tropical urban environment. PLoS Neglected Tropical Diseases 4, e920.
  • Wan NO, Chen CD, Soh HN et al. (2009) Surveillance of Aedes mosquitoes in a university campus in Kuala Lumpur, Malaysia. Tropical Biomedicine 26, 206215.
  • Williams CR, Long SA, Russell RC et al. (2006) Optimizing ovitrap use for Aedes aegypti in Cairns, Queensland, Australia: effects of some abiotic factors on field efficacy. Journal of the American Mosquito Control Association 22, 635640.
  • Williams CR, Bader CA, Kearney MR et al. (2010) The extinction of dengue through natural vulnerability of its vectors. PLoS Neglected Tropical Diseases 4, e922.
  • World Health Organization (WHO) (2012a) Dengue. Retrieved 14 June 2012, from http://www.who.int/denguecontrol/en/.
  • World Health Organization (WHO) (2012b) Dengue and severe dengue: factsheet no. 117. Retrieved 21 June 2012, from http://www.who.int/mediacentre/factsheets/fs117/en/.
  • Wu PC, Guo HR, Lung SC et al. (2007) Weather as an effective predictor for occurrence of dengue fever in Taiwan. Acta Tropica 103, 5057.