Climate change and dengue: a critical and systematic review of quantitative modelling approaches
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Naish et al. BMC Infectious Diseases 2014, 14:167
http://www.biomedcentral.com/1471-2334/14/167
RESEARCH ARTICLE Open Access
Climate change and dengue: a critical and
systematic review of quantitative modelling
approaches
Suchithra Naish1*, Pat Dale2, John S Mackenzie3, John McBride4, Kerrie Mengersen5 and Shilu Tong1
Abstract
Background: Many studies have found associations between climatic conditions and dengue transmission.
However, there is a debate about the future impacts of climate change on dengue transmission. This paper
reviewed epidemiological evidence on the relationship between climate and dengue with a focus on quantitative
methods for assessing the potential impacts of climate change on global dengue transmission.
Methods: A literature search was conducted in October 2012, using the electronic databases PubMed, Scopus,
ScienceDirect, ProQuest, and Web of Science. The search focused on peer-reviewed journal articles published in
English from January 1991 through October 2012.
Results: Sixteen studies met the inclusion criteria and most studies showed that the transmission of dengue is
highly sensitive to climatic conditions, especially temperature, rainfall and relative humidity. Studies on the potential
impacts of climate change on dengue indicate increased climatic suitability for transmission and an expansion of
the geographic regions at risk during this century. A variety of quantitative modelling approaches were used in the
studies. Several key methodological issues and current knowledge gaps were identified through this review.
Conclusions: It is important to assemble spatio-temporal patterns of dengue transmission compatible with long-term
data on climate and other socio-ecological changes and this would advance projections of dengue risks associated with
climate change.
Keywords: Climate, Dengue, Models, Projection, Scenarios
Background these are potentially associated with substantial increases
Dengue is a major public health concern for over half of in dengue outbreaks. Apart from climate factors other
the world’s population and is a leading cause of hospital- important issues that potentially contribute to global
isation and death, particularly for children in endemic changes in dengue incidence and distribution include
countries [1]. Recent studies estimate that 3.6 billion population growth, urbanisation, lack of sanitation, in-
people are at risk, with over 230 million infections, creased human travel, ineffective mosquito control, and
millions of cases with dengue fever, over 2 million cases increased reporting capacity [3-9].
with severe disease, and 21,000 deaths [1-5]. A 30-fold Dengue is primarily transmitted by Aedes aegypti and
increase in the number of dengue cases over the past secondarily by Aedes albopictus. Both mosquitoes have
50 years has been recorded with nearly 119 countries en- adapted to local human habitation with oviposition and
demic for dengue [4]. As part of global climate changes, larval habitats in natural (e.g., rock pools, tree holes and
temperature has increased by a global average of 0.75°C leaf axis) and artificial (e.g., water tanks, blocked drains,
over the past 100 years. Temperature increases such as pot plants and food and beverage containers) collections
in the urban and peri-urban environment. These mosqui-
* Correspondence: s.naish@qut.edu.au toes may be infected with any of the four dengue viruses
1
School of Public Health and Social Work & Institute of Health and
Biomedical Innovation, Queensland University of Technology, Victoria Park
with an incubation period of 3–14 days [10]. Dengue
Road, Brisbane, Queensland, Australia affects people of all ages, including infants irrespective of
Full list of author information is available at the end of the article
© 2014 Naish et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly credited.Naish et al. BMC Infectious Diseases 2014, 14:167 Page 2 of 14
http://www.biomedcentral.com/1471-2334/14/167
gender. Dengue viruses cause a spectrum of disease, with better understanding of the relationship between climate
symptoms from mild influenza-like symptoms to severe and disease is an important step towards finding ways
or fatal haemorrhage fever [11]. to mitigate the impact of disease on communities, for
example, malaria [39]. Successful future management of
The ecology of dengue and vector dengue requires an understanding of the dynamics of the
The epidemiological triangle of dengue includes host, virus, host, vector, and environmental factors especially
pathogen and mosquito vectors (including Ae. aegypti in the context of a changing climate [21].
and Ae. Albopictus) together with their interactions in
the environment. Dengue is climate sensitive as the virus Quantitative modelling of dengue and climate
has to complete part of its development in the mosquito The influences drawn about the relationships between
vectors that transmit the disease [5]. The major vector is dengue and climate, and the predictions of dengue
Ae. aegypti whose life cycle is directly influenced by ambi- under future climate change scenarios may depend on
ent temperature and rainfall [12]. Increased temperature the analytical approaches used. The aim of this paper is
could increase dengue risk by increasing the rate of mos- to review the relevant literature on dengue disease and
quito development and reducing virus incubation time in climate with a focus on quantitative models of the impact
areas where the vector presently exists, thereby increasing of climate change; to address methodological issues in
the rate of transmission [13-16]. Conversely, extreme hot this challenging field and then to indicate future oppor-
temperatures may also increase the rate of mosquito mor- tunities and research directions.
tality and thus decrease dengue risk [17]. Similarly, rainfall
can have non-linear contrasting effects on dengue risk Methods
[6,17]. Heavy rainfall may flush away eggs, larvae, and Search strategy
pupae from containers in the short term but residual A literature search was conducted in October 2012 using
water can create breeding habitats in the longer term [18]. the electronic databases PubMed, Scopus and Science-
A dry climate can lead to human behaviour of saving Direct, ProQuest and Web of Science to obtain the infor-
water in water storage containers, which may become mation on the impact of climate variables (and climate
breeding sites for Ae. Aegypti [19]. Thus, climatic condi- change) on dengue disease transmission. The search period
tions may affect the virus, the vector and/or human be- included January 1991 (the commencement of the report-
haviour both directly and indirectly [20]. Studies have ing of dengue in Queensland, Australia) to October 2012.
demonstrated that the ecology of virus is intrinsically tied We limited the literature search to journal articles pub-
to the ecology of dengue vectors [21]. Recent studies have lished in English, available in full-text/ pdf. The key words
also elaborated on the impacts of climate change on the used were dengue disease transmission, climate and/or cli-
vector, for example, how the extreme climatic events drive mate change, projection /forecast and scenario but not
mosquito outbreaks [22-25]. However, empirical relation- diagnostic tests, vector/ virus type. References and citations
ships have been demonstrated between climate variables, of the articles identified were checked to ensure that all
dengue and dengue vectors, casual relationships have not relevant articles were included (Figure 1).
been strongly established.
Selection criteria
Dengue and climate change Three selection criteria were used to select articles from
It is established that climate change is happening and it the search results for the detailed consideration. First,
is likely to expand the geographical distribution of sev- in order to obtain authoritative information, this review
eral mosquito-borne diseases [26]. The mounting evi- included only peer-reviewed journal articles. Second, ar-
dence around climate-disease relationships raises many ticles had to include larger geographical areas, climate
important issues about the potential effects of global cli- data including climate parameters and statistical analytical
mate changes on the transmission of infectious diseases, methods and at least one climate-based projection of fu-
particularly dengue [5,27-29]. There is evidence indicat- ture dengue disease transmission. Finally, we included only
ing that dengue epidemics have been associated with quantitative studies based on statistical models (climate-
temperature [30-32], rainfall [33,34] and relative humidity based) because qualitative studies used an entirely different
[35-37]. Few studies have included spatial data in climate- research design and analytic approach.
based predictive models [33,38]. We assessed the strengths and limitations of analytical
As global climate change is predicted to accelerate models and their use of established relationship between
over the next a few decades at least [1,3-6,27], an in- climate change and dengue transmission. Finally, we pro-
creased frequency, intensity and duration of extreme vide recommendations for future research directions to-
climatic events are more likely, so affecting dengue wards model parameters and predictions of climate change
transmission. This is a global public health priority. A on dengue transmission.Naish et al. BMC Infectious Diseases 2014, 14:167 Page 3 of 14
http://www.biomedcentral.com/1471-2334/14/167
Records identified through Additional records identified
Identification
database searching through other sources
(n = 551) (n = 34)
Records after duplicates removed
(n = 561)
Screening
Records screened Records excluded
(n = 531) (n = 456)
Full-text articles Full-text articles
assessed for eligibility excluded, with reasons
Eligibility
(n = 75) (n = 59)
Studies included in
qualitative synthesis
(n = 16)
Included
Figure 1 Flowchart of literature search.
Results stations. With the exception of a few studies that used
The initial search yielded 531 studies of which 456 global gridded projection data sets [34,38], all other stud-
were deemed to be potentially relevant and were sub- ies obtained climate variables from local/ national me-
jected to further perusal. This led to 75 studies con- teorological stations [31,32,36,40-43,45,49,50]. Data on
sidered in detail and 16 that strictly met the inclusion socio-economic heterogeneity, climatic diversity includ-
criteria. Table 1 shows the characteristics of the 16 studies ing both tropical and subtropical, and un-observed con-
that addressed the climate variables (climate change) and founders such as social behaviour were described in
future risk of dengue transmission, using climate change individual studies.
scenarios.
Analytical approaches
Dengue data Among the selected studies, six used SARIMA- time
For the purpose of this study, we have considered Dengue series or wavelet time series models [34,36,40,42,45,50],
Fever (DF) or Dengue Hemorrhagic Fever (DHF) as a four used different types of regression analysis [9,32,33,41]
single entity ‘dengue’ as such kind of information was and other studies have variously employed a general
not available from many studies. Eight studies used additive mixed (GAM) model [30], a spatial model [43],
monthly confirmed and reported dengue notified case data a non-linear model [31], a multivariate model [46], and
[8,31,33,34,40-43], two studies used annual cases [44,45], global circulation models (GCM) towards projections
three studies used weekly confirmed case data [9,32,36] [33,38]. Most studies used a combination of the different
aggregated from daily surveillance and one study used analytical models but to varying degrees. For example,
daily cases [30]. Two studies combined dengue-specific among studies that used regression analysis, some intro-
(entomological) parameters [38,46]. Most of the studies duced autoregressive terms. A few studies used Poisson
used dengue laboratory-confirmed case data (notified) ob- regression models to allow for autocorrelation and over
tained from health departments (Table 1) and some used dispersion. A number of studies incorporated the prob-
reported cases [36,47,48]. ability of risk of seasonal forecasts as covariates in their
analysis [31,36] and only few studies demonstrated the
Covariate data potential use of forecasting in the development of climate-
Most data were aggregated to monthly estimates from daily, driven models [33,38,44]. The details of these models are
weekly and annual data obtained from meteorological explained below.http://www.biomedcentral.com/1471-2334/14/167
Naish et al. BMC Infectious Diseases 2014, 14:167
Table 1 Studies included in the review of climate impact on dengue
References Study area Dengue data Covariate data Spatial resolution Analytical approaches Key findings Comments
(Period)
Earnest et al. [32] Singapore Weekly laboratory Weekly climate (mean/ Local meteorological Poisson regression, Temperature, relative humidity Temporal trends of dengue
(2001–2008) confirmed notified minimum/maximum station data Sinusoidal function and SOI associated with dengue were noticeable.
dengue cases temperature, mean rainfall, cases.
mean/minimum/maximum
relative humidity, mean
hours of sunshine and
mean hours of cloud) data
Descloux et al. Noumea Monthly confirmed Monthly climate (temperature, Local meteorological Non-linear models Significant inter-annual The epidemic dynamics of
[31] (New Caledonia) cases of DF/ DHF precipitation, relative humidity, station data correlations were observed dengue were driven by
(1971–2010) wind force, potential between dengue outbreaks climate.
evapo-transpiration, hydric and summertime temperature,
balance sheet) data and precipitation, relative humidity
ENSO indices but not ENSO.
Chen et al. [30] Taiwan Daily confirmed Daily climate (temperature, Local meteorological GAM Rainfall was correlated with A climatic change does have
(1994–2008) cases of notified DF rainfall) data, socio- station data dengue cases. Lag effects were influence on dengue
demographic factors observed. outbreaks.
Hu et al. [43] Australia Monthly confirmed Monthly weather, SEIFA, pop Local meteorological Bayesian CAR Increase in dengue cases of 6% Socio-ecological factors
(2002–2005) cases of notified (LGA) station data in association with a 1-mm appear to influence dengue.
dengue increase in average monthly The drivers may differ for
rainfall and a 1°C increase in local and overseas cases.
average monthly maximum Spatial clustering of dengue
temperature, respectively was cases was evident.
observed.
Chowell [45] Peru Annual confirmed Time series of annual Local meteorological Wavelet time series A significant difference in the The differences in the timing
(1994–2008) cases population size and density, station data timing of epidemics between of dengue epidemics
altitude and climate data jungle and coastal regions was between jungle and coastal
observed. regions were significantly
associated with the timing of
the seasonal temperature
cycle.
Thai et al. [50] Vietnam Monthly confirmed Monthly climate (mean Local meteorological Wavelet time series ENSO indices and climate Climate variability and ENSO
(1994–2009) cases temperature, rainfall and station data variables were significantly impact dengue outbreaks.
relative humidity) data associated with dengue
and ENSO indices incidence.
Colon-Gonzalez Mexico Monthly confirmed Monthly climate (minimum Local meteorological Linear regression, Incidence was higher during Temperature was an
et al. [41] (1985–2007) cases and maximum temperature station data Phillips–Perron and El-Nino. Incidence was important factor in the
and rainfall) and ENSO Jarque–Bera test associated with El-Nino and dengue incidence in Mexico.
indices tests temperature during cool and
dry times.
Pinto et al. [9] Singapore Weekly confirmed Weekly climate (maximum Local meteorological Poisson regression, For every 2–10 degrees of Temperature was the best
(2000–2007) notified DF cases and minimum temperature, station data Principal component maximum and minimum predictor for the dengue
maximum and minimum anlaysis temperature variation, an increase in Singapore.
Page 4 of 14
relative humidity) data increase of cases of 22-184%
and 26-230% respectively, was
observed.Table 1 Studies included in the review of climate impact on dengue (Continued)
http://www.biomedcentral.com/1471-2334/14/167
Naish et al. BMC Infectious Diseases 2014, 14:167
Gharbi et al. [36] French West Weekly laboratory Weekly climate (cumulative Local meteorological Time series (SARIMA), Temperature was significantly Temperature improves
Indies confirmed cases rainfall, relative humidity, station data RMSE and Wilcoxon associated with dengue dengue outbreaks better
(2000–2007) from hospitals or minimum, maximum and signed-ranks test forecasting but not humidity. than humidity and rainfall.
not average temperature) data
Hu et al. [42] Australia Monthly confirmed Monthly SOI, rainfall and Local meteorological Cross-correlations, A decrease in the SOI was Climate variability is directly
(1993–2005) cases of notified DF annual population station data Time series (SARIMA) significantly associated with an and/or indirectly associated
cases increase in the dengue cases. with dengue. SOI based
epidemic forecasting is
possible.
Johansson et al. Puerto Rico, Monthly reported Monthly climate (precipitation, Global climate Wavelet time series Temperature, rainfall and dengue The role of ENSO may be
[48] Mexico, Thailand cases of DF/ DHF minimum, maximum and surfaces (0.5 × 0.5°) incidence were strongly obscured by local climate
(1986–2006) mean average temperature) local meteorological associated in all three countries heterogeneity, insufficient
data and ENSO indices station data for the annual cycle. The data, randomly coincident
associations with ENSO varied outbreaks, and other,
between countries in the potentially stronger, intrinsic
multi-annual cycle. factors regulating dengue
transmission dynamics.
Bambrick et al. Australia Annual incidence – Annual Temperature, vapour Local meteorological Climate change Geographic regions with An eight-fold increase in
[44] (1991–2007) notified cases of DF pressure and population station data scenarios climates that are favourable the number of people living
to dengue could expand to in dengue prone regions in
include large population centres Australia will occur unless
in a number of currently greenhouses gases are
dengue-free regions. reduced.
Bulto et al. [46] Cuba (1961–1990) Dengue-specific Monthly climate (maximum Local meteorological Multivariate (Empiric Strong associations between Climate variability has
parameters of and minimum temperature, station data orthogonal function) climate anomalies and dengue influence on dengue.
DF/ DHF precipitation, atmospheric
pressure, vapour pressure,
relative humidity, thermal
oscillation and solar radiation)
data
Cazelles et al. [40] Thailand Monthly confirmed Monthly climate (temperature Local meteorological Wavelet time series Strong association between The association is
(1983–1997) cases of DHF and rainfall) data and ENSO station data dengue incidence and El-Nino non-stationary and have
indices events was observed. a major influence on the
Temperature had greater synchrony of dengue
influence on dengue than epidemics.
rainfall.
Hales et al. [33] Global Monthly reported Monthly climate (maximum, Region-specific and Logistic regression In 2085, under climate and There is a potential increase
(1975–1996) cases of DF minimum and mean GCM projections and IPCC scenarios population projections, 50-60% in the dengue risk areas
temperature, rainfall and of the population would be at under climate change
vapour pressure) and dengue risk. scenarios, if the risk factors
population and projections remain constant.
(future climate and
population) data
Page 5 of 14Table 1 Studies included in the review of climate impact on dengue (Continued)
http://www.biomedcentral.com/1471-2334/14/167
Naish et al. BMC Infectious Diseases 2014, 14:167
Patz et al. [38] Global Dengue-specific Monthly climate data Site-specific GCM GCM output to Among the three GCMs, the Epidemic potential increased
(1931–1980) parameters vectorial capacity average projected temperature with a relatively small
elevation was 1.16°C, expected temperature rise, indicating
by the year 2050. that lower mosquitoes
infestation values would be
necessary to maintain or
spread dengue in a
vulnerable population.
Page 6 of 14Naish et al. BMC Infectious Diseases 2014, 14:167 Page 7 of 14
http://www.biomedcentral.com/1471-2334/14/167
Linear regression models interannual variation in local or regional climate linked to
Colon-Gonzalez et al. [41] used multiple linear regression El-Niño may determine the temporal and spatial dynamics
models to examine the associations between changes in of dengue. Thai et al. [50] used a wavelet time series ana-
the climate variability and dengue incidence in the warm lysis to investigate the associations between climate vari-
and humid regions of Mexico for the years 1985–2007. ables including mean temperature, humidity and rainfall,
Their results showed that the incidence was higher dur- and ENSO indices and dengue incidence in Vietnam
ing El- Niño events and in the warm and wet season. during the period 1994 to 2009. Their results showed
Their study demonstrated that dengue incidence was that the ENSO indices and climate variables were sig-
positively associated with the strength of El-Niño and the nificantly associated with dengue incidence in the 2 to
minimum temperature, especially during the cool and dry 3-year periodic band, although the associations were
season. transient in time. Chowell [45] used wavelet time series
analysis to determine the relationship between climatic
Time series/wavelet time series models factors including mean, maximum and minimum tem-
Time series modelling approaches have been extensively perature and rainfall and dengue incidence for the period
applied in assessing the impact of climate variables on 1994–2008 in jungle and coastal regions of Peru. They
dengue incidence. For example, Gharbi et al. [36] fit- revealed that incidence was highly associated with sea-
ted a seasonal autoregressive integrated moving aver- sonal temperature and suggested that dengue was fre-
age (SARIMA) model of dengue incidence and climate quently imported into coastal regions through infective
variables including temperature, rainfall and relative sparks from endemic jungle areas and/or cities of other
humidity in French West Indies for the period 2000– neighbouring endemic countries.
2006. They found that temperature significantly improved
the ability of the model to forecast dengue incidence Poisson regression models
but this was not so for humidity and rainfall. They also Poisson regression models have been applied in deter-
found that minimum temperature at 5 weeks lag time mining the relationship between climate and dengue.
was the best climatic variable for predicting dengue For example, Earnest et al. [32] used this approach
outbreaks. Similarly, Hu et al. [42] used a time series to determine the association between climate variables
SARIMA model to examine the impact of El-Niño on (temperature, humidity, rainfall), ENSO indices and den-
dengue in Queensland, Australia for the period 1993– gue in Singapore. They found that temperature, relative
2005. They suggested that a lower Southern Oscillation humidity and ENSO were significantly and independently
Index (SOI) was related to increased dengue cases. associated with dengue cases. No one set of climate vari-
Wavelet time series analysis has been applied to exam- ables was superior to the others, so they suggested that
ine the associations between El-Niño Southern Oscillation all the climate variables had a similar predictive ability.
(ENSO), local weather, and dengue incidence in Puerto Pinto et al. [9] used Poisson regression model to deter-
Rico, Mexico, and Thailand [34] particularly, with the aim mine the impact of climate variables (temperature, rain-
of identifying time- and frequency-specific associations. In fall and relative humidity) on dengue cases in Singapore.
all three countries, temperature, rainfall, and dengue inci- They found that for every 2-10°C of variation of the
dence were strongly associated on an annual scale. On a maximum temperature, dengue cases were increased by
multiyear scale, ENSO was associated with temperature 22.2-184.6%. For the minimum temperature, for the same
and with dengue incidence in Puerto Rico, but only for variation, they observed that there was an average in-
part of the study period. Only local rainfall was associated crease of 26.1-230.3% in the number of the dengue
with the incidence of dengue in that country. The lack of a cases from April to August. Their study concluded that
direct association between ENSO and weather variables to the variable temperature (maximum and minimum) was
dengue incidence suggests that the ENSO-dengue associ- the best predictor for the increased number of dengue
ation may be a spurious result. In Thailand, ENSO was as- cases.
sociated with both temperature and rainfall, and rainfall Chen et al. [49] applied Poisson regression using a
was associated with dengue incidence. However, detailed GAM model to examine the relationship between pre-
analysis suggested that this latter association was also cipitation and dengue in Taiwan for the period 1994–
probably spurious. The authors concluded that there was 2008. The GAM allows a Poisson regression to be fit as
no significant association between any of the variables in a sum of nonparametric smooth functions of predictor
Mexico on the multiyear scale. In another study, Cazelles variables. They found that differential lag effects follow-
et al. [40] used a wavelet time series analysis to demon- ing precipitation were statistically associated with in-
strate a strong non-stationary association between dengue creased risk of dengue. Poisson regression, using a GAM
incidence and El-Niño in Thailand for the years 1986 model was used to evaluate the multiple-lag effects of
to 1992. They suggested that under certain conditions, stratified precipitation levels on specific diseases. AllNaish et al. BMC Infectious Diseases 2014, 14:167 Page 8 of 14
http://www.biomedcentral.com/1471-2334/14/167
models were adjusted for the multiple-lag effects of daily (except for [49]) followed by rainfall in the majority of
temperature, month, and township for evaluating the as- research.
sociations between categorized extreme precipitation and
diseases, with further trend tests performed to examine Projections of climate change impacts on dengue
linear associations between levels of precipitation and Patz et al. [38] examined the potential risk posed by global
outbreaks of each disease. climate change on dengue transmission. They used vector-
ial capacity equation that was modified to estimate the epi-
Bayesian models demics of dengue. The model used project climate change
Bayesian spatial conditional autoregressive modelling data from global circulation model (GCM) at 250 km x
approaches have been used to demonstrate the impact 250 km resolution project future risk of dengue globally.
of climatic, social and ecological factors on dengue in Their findings suggest that increased incidences have pre-
Queensland, Australia [43]. The authors suggested that dominantly occurred in regions bordering endemic zones
6% increase in locally acquired dengue was observed in in latitude or altitude. They found that epidemic activity
association with a 1-mm increase in average monthly increased with a small rise in temperature, indicating
rainfall and a 1°C increase in average monthly maximum that fewer mosquitoes would be necessary to maintain or
temperature. They also reported that overseas-acquired spread dengue in a population at risk of dengue. They con-
dengue cases were increased by 1% in association with a cluded that transmission may be saturated in hyper-
1-mm increase in average monthly rainfall and a 1-unit endemic tropical regions and human migration patterns of
increase in average socioeconomic index, respectively. susceptible individuals are likely to be more important to
overall transmission than are climatic factors. Endemic lo-
Non-linear models cations may be at higher risk from dengue if transmission
Descloux et al. [31] developed an early warning system intensity increases. Hales et al. [33] estimated the changes
using a long-term data set (39 years) including dengue in the geographical limits of dengue transmission from
cases and meteorological data (mean temp, min and 1975 to 1996 and the size of population at risk, using logis-
max temperature, relative humidity, precipitation and tic regression. Monthly averages of vapour pressure, rain-
ENSO indices) in New Caledonia, using multivariate fall, and temperature recorded between 1961 and 1990,
non-linear models. They observed a strong seasonality with or without statistical interaction terms between vari-
of dengue epidemics and found significant inter-annual ables were included in their statistical models. Based on
correlations between epidemics and temperature, pre- data from GCM projections and human demographics,
cipitation and relative humidity. Bulto et al. [46] applied they predicted that dengue would increase and include a
empiric orthogonal function (EOF) to estimate the asso- larger total population and higher percent of the popula-
ciation between climate data including monthly max- tion. Although their studies suggested that the dengue dis-
imum and minimum mean temperatures, precipitation, tribution was climate dependent, other factors needed to
atmospheric pressure, vapour pressure, relative humid- be considered in addition to climate during epidemics.
ity, thermal oscillation, days with precipitation, solar Bambrick et al. [44] highlighted the potential for cli-
radiation in, and isolation and dengue in Cuba during mate change to affect the safety and supply of blood glo-
the period 1961–1990. The EOF is designed to obtain bally through its impact on vector-borne disease, using
the dominant variability patterns from sets of fields of the example of dengue in Australia as a case-study. They
any type, synthetic indicators or indexes, and summar- modelled geographic regions that were suitable for den-
ise the variability observed in a group of variables. Cli- gue transmission over the coming century under four
matic anomalies included were multivariate ENSO index, climate change scenarios, estimated changes to the pop-
quasi-biennial oscillation and North Atlantic Oscillation. ulation at risk and effect on blood supply. They applied
They found a strong association between climate anom- logistic regression models using climate change scenar-
alies and dengue which demonstrated significant climate ios to the observed geographic distribution of dengue in
variability. Australia. The most important predictor variables in the
In summary, the quantitative models employed for models were vapour pressure and temperature. Their re-
evaluating the relationship between climate variables and sults indicated that geographic regions with climates that
dengue have been typically different with respect to the are favourable to dengue transmission could expand to
distributional assumptions (e.g., normal, Poisson), the na- include large population centres in a number of cur-
ture of the relationship (linear and non-linear) and the rently dengue-free regions in Australia.
spatial and/or temporal dynamics of the response. Overall,
the models consistently reveal variability in the relationship Methodological issues
between dengue and climate variables, related to country, Several methodological issues emerged when we con-
but the methods identified an association with temperature ducted this literature review on climate change andNaish et al. BMC Infectious Diseases 2014, 14:167 Page 9 of 14
http://www.biomedcentral.com/1471-2334/14/167
dengue. These issues include the study design, analytical dengue [34,40,50]. Although annual periodic patterns
model, time period, scale of analysis, exposure variables are a common phenomenon in dengue endemic areas,
and other factors associated with dengue transmission as the identification of a periodic multi-annual (e.g., 2 to
illustrated below. 3 years) cycle differs between countries as well as in ana-
lyses used. Cazelles et al. [40] used wavelet approaches to
Study designs demonstrate a highly significant but discontinuous associ-
Several study designs have been considered by various ation between ENSO, precipitation and dengue epidemics
authors while studying the relationships between climate in Thailand.
and dengue (Table 1). For example, Hales et al. [47] used A continuous annual mode of oscillation with a non-
a mixed ecological study design and long-term data to stationary 2 to 3-year multi-annual cycle was found with
determine the relationship between the annual number strong irregular associations between dengue incidence
of dengue cases, ENSO, temperature and rainfall using and ENSO indices and climate variables in Vietnam [50].
global atmospheric analyses climate-based data. Descloux Although these wavelet analyses have provided important
et al. [31] used long-term observational data and examined contribution to the cyclical dynamics of dengue transmis-
inter-annual correlations between ENSO, local climate sion, the associations with ENSO have been irregular and
and dengue. temporary, which reduce the potential for estimating fu-
ture predictions based on these climate anomalies.
Analytical models
Time series modelling approaches have been applied to Time period
estimate the baseline relationships between climate and Choosing a baseline time period for climate data is also
dengue [9,36,40,42,50,51] (Table 1). SARIMA models are important. Climate and dengue relationships in the same
potentially useful when there are time dependences be- city can be very different between the 1960s and the
tween each observation [36,52]. The assumption that 2010s. Differences could be due to socioeconomic and,
each observation is correlated to previous ones makes it demographic changes and urbanisation. Differences in
possible to model a temporal structure, with more reli- the time periods used to estimate the historical climate
able predictions, especially for climate-sensitive diseases and dengue relationships also make it difficult to com-
(e.g., mosquito-borne diseases), than those obtained by pare projections across studies. Therefore, it has been
other statistical methods. SARIMA models have been recommended that long-term (generally, at least many
successfully used in epidemiology to predict the evolu- decades) baseline climate data be used for modelling
tion of infectious diseases. Moreover, these models allow climate-based diseases to calculate an average that is not
the integration of external factors, such as climatic vari- influenced by climate variability [54,55].
ables, that may increase the predictive power and ro-
bustness of predictive models due to longer time periods Spatial and temporal scales
of data. Another issue to be considered when modelling is the
Some studies have used Fourier analysis to analyse rela- spatio-temporal scale of analysis. This is because spatial
tionships between oscillating time series. This technique and temporal characteristics may provide useful infor-
decomposes time series into their periodic components mation on risk assessments to be used by local or national
that can then be compared between time series. Since dengue prevention and control programs to prepare for
Fourier analysis cannot take into account temporal and respond to dengue epidemics in endemic settings.
changes in the periodic behaviour of time series (i.e. non- Dengue may be sensitive to differences in climatic con-
stationarity), this method, and others such as generalized ditions at a local, regional or global level. At the global
linear models, may be inadequate for investigating the de- level, there is the potential for climate-related spread of
terminants of transmission dynamics of dengue [53]. dengue. There are areas that are currently only at risk
Wavelet analysis is suitable for investigating time and not endemic for dengue, but may become endemic
series data from non-stationary systems and for infer- as climate changes, especially related to temperature
ring associations between such systems [53]. This ap- change. For example, Patz et al. [38] used a process-
proach reveals how the different scales (i.e. the periodic based model considering an entomological variable, i.e.,
components) of the time series change over time. Wavelet the vectorial capacity (VC). They included temperature
analysis is able to measure associations between two time effects over VC and predicted the potential global den-
series at any period. Wavelet analyses have been used gue spread using climate change scenarios for 2050.
to compare time series of disease incidence across lo- Therefore, we suggest to develop models at a local and/or
calities and countries for the characterisation of the regional scale that can increase their predictive capacity
evolution of epidemic periodicity and the identification of [56] and incorporate important local factors that affect dis-
synchrony. Wavelet analyses have been used in analysing ease transmission [54].Naish et al. BMC Infectious Diseases 2014, 14:167 Page 10 of 14
http://www.biomedcentral.com/1471-2334/14/167
Climate variables should also be considered. For example, Gubler et al. [63]
The choice of climate variables in the modelling is claimed that a dramatic urban growth has occurred in the
an important issue to consider. The following climate past 40 years providing the suitable ecological conditions
variables have been considered in the studies: max- for Ae. aegypti to increase in close association with large
imum, minimum and mean temperature, rainfall, relative and crowded human populations in tropical areas, creating
humidity, ENSO indices and sunshine. Among these, ideal conditions for dengue transmission.
temperature and ENSO indices have been found to be Dengue-infected human movement should be consid-
important in any study. Authors have identified that ered as another important factor [64] since Aedes gen-
other variables such as river levels are also included along erally has a short flight range so is unlikely to spread
with climate variables [57], however, these studies are out dengue over large distances. For example, Rabbai et al.
of the study scope. Hales et al. [33] developed an empir- [65] have demonstrated dengue viral exchange between
ical model fitting logistic regression and modelled dengue the urban areas of Ho Chi Minh City and other prov-
outbreak (presence/absence) considering climate baseline inces of southern Vietnam and suggested that human
data for the period 1961–1990. They found that the movement between urban and rural areas may play a key
vapour pressure was the only explanatory climate vari- role in the transmission of dengue virus across southern
able responsible for global dengue transmission. Using Vietnam.
this model, applying climate projections for 2085 from a
GCM, Hales et al. predicted a limited extension of den- Challenges
gue. In both the projection studies, climate projections The fundamental challenge for predicting dengue trans-
were based on historical exposure-response functions of mission is how to best model future climate at a regional
climate variables and dengue incidence that are applied and/or local level. In another words, how can we appro-
to climate change models and emissions scenarios to esti- priately downscale the GCM modelling outcomes to a
mate and predict future dengue distribution. Therefore, regional and/or local level? The Intergovernmental Panel
it is important to consider which climate variables are on Climate Change (IPCC) has developed 40 Special
the best predictors of dengue transmission and the re- Reports on Emissions Scenarios (SRES) covering a wide
search reported here indicates that temperature is con- range of main driving forces of future green house gas
sistently important but that vapour pressure or relative emissions [26]. These scenarios were categorised into
humidity are also significant contributors. four classes: A1, A2, B1 and B2. A1 characterises rapid
economic growth, population growth by 2050, introduc-
Other factors associated with dengue transmission tion of new and efficient technologies. A2 characterises
There are other environmental, socio-economic and demo- high population growth, slow economic development
graphic factors associated with dengue transmission and and technological changes. B1 characterises the similar
the relative contribution of these factors may differ be- population growth like A1 but with rapid economic
tween settings (scales, countries, regions). These include, changes. B2 characterises medium population and eco-
but are not limited to attenuators (e.g., mosquito manage- nomic growth with localised solutions to economic,
ment, screening dwellings, using personal insect repellents social and environmental sustainability. These scenar-
and bed nets) or exacerbators (e.g., actively storing of water ios can be used to project future climate based on GCM
in open containers, passively allowing water to remain in models [26].
bins, garden accoutrements) [58]. These are outside the Selecting climate models is not a trivial task, con-
scope of our current review. However, socio-demographic sidering the strengths and limitations of various GCM
change is important and we have briefly considered this models. The IPCC recommended that no single GCM
below. can be considered the best and that various GCMs
should be applied to account for climate modelling un-
Associations between socio-demographic changes certainties [26]. These do not necessarily indicate errors
and dengue in the modelling and should therefore not be used to
Other etiologic factors that impact dengue transmission conclude that a model is inaccurate. As the global tem-
include social and demographic changes, economic sta- perature increases, tropical insects may spread their habi-
tus, human behaviour and education [7,43,59-61]. Reiter tats into more northern or southern latitudes which can
et al. [62] evidenced that dengue virus transmission was result in higher transmission. This intriguing idea was
limited by human life style in Texas. Other important first suggested by Shope [12] and was considered further
human related factors such as exceptional population by other researchers [15,66]. However, there is still de-
growth, unplanned urbanisation and air travel needs to be bate whether the increased pattern of dengue arose from
considered in modelling studies [63]. Other factors driven climate change [16] or from socio-economic changes
by economic growth such as animals and commodities in combination with ecologic and demographic changesNaish et al. BMC Infectious Diseases 2014, 14:167 Page 11 of 14
http://www.biomedcentral.com/1471-2334/14/167
[63,67]. Finally, other etiological factors must ultimately rainfall followed by low/ lack of rainfall could intimi-
be incorporated into integrated modelling to determine date sharp decrease in mosquito populations, for example,
human risk to dengue [68]. in Puerto Rico.
Dengue transmission in endemic settings is charac-
Discussion terised by non-linear dynamics, with strong seasonality,
Climatic factors play a significant role in the mosquito multi-annual oscillations and non-stationary temporal
biology, the viruses they transmit, and more broadly, variations. Seasonal and multi-annual cycles in dengue
dengue transmission cycles. Higher temperatures in- incidence vary over time and space. Besides the sea-
crease the rate of larval development and shorten the sonality of dengue transmission, periodic epidemics and
emergence of adult mosquitoes, increase the biting rate more irregular intervals of outbreaks are commonly
of mosquito and reduce the time required for virus rep- observed [34,75-77].
lication within the mosquito. Extreme higher tempera- Evidence suggests that inter-annual and seasonal
tures may reduce mosquito survival time, which could climate variability have a direct influence on the transmis-
offset the positive effect on mosquito abundance [69]. sion of dengue [17,32,34,41,78,79]. This evidence has been
Evidence had accrued to show the link between tem- assessed at the country level in order to determine the pos-
perature and dengue incidence rates [9,34,36,47]. These sible consequences of the expected future climate change
studies have used a range of statistical approaches con- [33,38]. These studies have highlighted that many climatic
sidering different temperature parameters (e.g. mean, variables play a key role in dengue transmission as dis-
maximum and minimum temperatures), and the results cussed above.
are generally consistent, indicating that the epidemics of Many studies have highlighted the importance of lag
dengue are driven by climate to some extent. time, at monthly scales. For example, in Taiwan, there
Relative humidity is another key factor that influences was a significant positive correlation with the maximum
mosquitoes’ life cycle at different stages. The combined temperature at lag 1–4 months, the minimum tem-
effect of temperature and humidity significantly influ- perature at lag 1–3 months and the relative humidity at
ences the number of blood meals and can also affect the lag 1–3 months [49]. In New Caledonia, significant asso-
survival rate of the vector, and the probability that it will ciations were found between temperature and dengue
become infected and able to transmit dengue [70]. In transmission at lag-0 [31].
the literature, relative humidity and temperature are the The delayed effect (or time lag) of climatic variables on
two most important variables with potential impact dengue incidence could be explained by climatic factors
on dengue transmission. Vapour pressure or relative which do not directly influence the incidence of dengue
humidity is affected by a combination of rainfall and but do so indirectly. This is through their effect on the life-
temperature and influences the mosquito lifespan and cycle dynamics of both vector and virus. This starts with
thus the potential for transmission of the virus. Hales et al. mosquito hatching, larval and pupal development, adult
found that annual average vapour pressure was the most emergence and virus amplification, incubation in humans
important climatic predictor of global dengue occurrence culminating in a dengue outbreak and results in a cumula-
[33]. Therefore, temperature, rainfall and relative hu- tive time lag [36,70]. Depending on the respective lag be-
midity are important determinants of the geographic tween the biological cycle or mosquito life-stage and the
limits within which dengue transmission can be expected clinical symptoms, the lag between climate data and inci-
to continue, primarily through their effects on the Aedes dence data will differ. The lag is expected to be shorter for
vector. Furthermore, within areas where minimum thresh- minimum temperatures that are usually associated with
olds of these climate parameters are sufficient to maintain adult mosquito’s mortality, longer for high relative humid-
dengue transmission, seasonal fluctuations in these pa- ity, both related to adult survival and hatching. On the
rameters will be important determinants of the duration other side, the mean temperature is involved in all bio-
and potentially the intensity of transmission. logical cycles of Ae. aegypti that take more time to influ-
Several studies have shown that the increased temper- ence the dengue incidence [5,36,78].
atures and relative humidity are determining factors in
predicting changes of the dengue transmission [31,32]. Strengths and limitations of studies
Contrasting, rainfall did not appear to play a significant Several statistical analytical methods have been used
role because many breeding sites of Ae. aegypti were to determine the relationship between climate variables
more dependent on human behaviour than on rainfall (and climate change) and dengue, including cross corre-
for their development and survival [71-73]. This might lations, Poisson, logistic and multivariate regression,
partly explain the lower impact of rainfall compared SARIMA-time series and wavelet time series (Table 1).
to other climatic variables on the dengue incidence. Many have been successful [31,34,41,49,50] in establishing
However, Chaves et al. [74] suggested that a series of climate and dengue relationships and developing predictiveNaish et al. BMC Infectious Diseases 2014, 14:167 Page 12 of 14 http://www.biomedcentral.com/1471-2334/14/167 models of dengue based on climate relationships. Mini- confounding effects of urbanisation, population growth mum, maximum and mean temperatures, relative humidity and tourism development are required to develop scenar- and rainfall were the most important climate variables that ios based on future projections of population growth and predict the dengue. However, these variables are predictive socio-economic development, including human behaviour. at specific lags of time. 5) There is a clear need for inter-disciplinary collabora- Remarkably, due to data constraints, modelling ana- tions with ecologists, sociologists, micro-biologists, bio- lyses were conducted using aggregated data over large statisticians and epidemiologists. Two areas need special spatial scales or long time periods [34]. Studies based on attention: one is in the area of climate modelling to ad- long time scales and large geographic areas may be un- dress issues of spatial and temporal scale and analytical successful in describing the influences that happen over methods, and the other relates to dengue incidence data daily or weekly periods and climate changes that occur quality control with the reporting agency (e.g., laboratories, at country level [45]. hospitals, health centres) addressing issues such as under- In general, the predictive power and model robustness reporting and misdiagnosis, dengue case definition, clinical would be better improved with large data over longer or lab-confirmed diagnosis, inpatient and outs report- periods. For example, Gharbi et al. [36] developed statis- ing, specific ages and/or dengue severity reported. From tical predictive models that were build-up on
Naish et al. BMC Infectious Diseases 2014, 14:167 Page 13 of 14
http://www.biomedcentral.com/1471-2334/14/167
Received: 15 February 2013 Accepted: 20 March 2014 26. Working Group II: Impacts, adaptation and vulnerability. [http://www.ipcc.
Published: 26 March 2014 ch/ipccreports/tar/wg2/index.php?idp=361].
27. Ramasamy R, Surendran SN: Global climate change and its potential
impact on disease transmission by salinity-tolerant mosquito vectors in
References coastal zones. Front Physiol 2012, 3:198.
1. Gubler DJ: The economic burden of dengue. Am J Trop Med Hyg 2012, 28. Caminade C, Medlock JM, Ducheyne E, McIntyre KM, Leach S, Baylis M,
86:743–744. Morse AP: Suitability of European climate for the Asian tiger mosquito
2. Beatty ME, Letson GW, Margolis HS: Estimating the global burden of Aedes albopictus: recent trends and future scenarios. J R Soc Interface
dengue. In Proceedings of the 2nd International Conference on Dengue and 2012, 1:1–11.
Dengue Haemorrhagic Fever Phuket, Thailand; 2008.
29. Sly PD: Health impacts of climate change and biosecurity in the Asian
3. Halasa Y, Shepard D, Zeng W: Economic cost of dengue in Puerto Rico.
Pacific region. Rev Environ Health 2011, 26:7–12.
Am J Trop Med Hyg 2012, 86:745–752.
30. Chen SC, Hsieh MH: Modeling the transmission dynamics of dengue
4. Global Alert and Response (GAR): Impact of dengue. [http://www.who.
fever: Implications of temperature effects. Sci Total Environ 2012,
int/csr/disease/dengue/impact/en/]
431:385–391.
5. Dengue and climate. [http://www.cdc.gov/dengue/entomologyEcology/
31. Descloux E, Mangeas M, Menkes CE, Lengaigne M, Leroy A, Tehei T,
climate.html]
Guillaumot L, Teurlai M, Gourinat AC, Benzler J, Pfannstiel A, Grangeon JP,
6. Githeko AK: Advances in developing a climate based dengue outbreak
Degallier N, De Lamballerie X: Climate-based models for understanding
models in Dhaka, Bangladesh: challenges & opportunities. Indian J Med
and forecasting dengue epidemics. PLoS Negl Trop Dis 2012, 6:e1470.
Res 2012, 136:7–9.
32. Earnest A, Tan SB, Wilder-Smith A: Meteorological factors and El Niño
7. Walker KR, Joy TK, Ellers-Kirk C, Ramberg FB: Human and environmental
Southern Oscillation are independently associated with dengue infections.
factors affecting Aedes aegypti distribution in an arid urban environment.
Epidemiol Infect 2012, 140:1244–1251.
J Am Mosq Control Assoc 2011, 27:135–141.
33. Hales S, de Wet N, Maindonald J, Woodward A: Potential effect of
8. Thai KTD, Anders KL: The role of climate variability and change in the
population and climate changes on global distribution of dengue fever:
transmission dynamics and geographic distribution of dengue. Exp Biol
an empirical model. Lancet 2002, 360:830–834.
Med (Maywood) 2011, 236:944–954.
34. Johansson MA, Cummings DAT, Glass GE: Multiyear climate variability and
9. Pinto E, Coelho M, Oliver L, Massad E: The influence of climate variables
dengue-El Niño Southern Oscillation, weather, and dengue incidence in
on dengue in Singapore. Int J Environ Health Res 2011, 21:415–426.
Puerto Rico, Mexico, and Thailand: a longitudinal data analysis. PLoS Med
10. Chevillon C, Failloux A-B: Questions on viral population biology to
2009, 6:e1000168.
complete dengue puzzle. Trends Microbiol 2003, 11:415–421.
11. Halstead SB: Dengue. Lancet 2007, 370:1644–1652. 35. Chakravarti A, Kumaria R: Eco-epidemiological analysis of dengue
12. Shope R: Global climate change and infectious diseases. Environ Health infection during an outbreak of dengue fever, India. Virol J 2005, 2:32.
Perspect 1991, 96:171–174. 36. Gharbi M, Quenel P, Gustave J, Cassadou S, La Ruche G, Girdary L, Marrama L:
13. Focks DA, Daniels E, Haile DG, Keesling JE: A simulation model of the Time series analysis of dengue incidence in Guadeloupe, French West Indies:
epidemiology of urban dengue fever: literature analysis, model forecasting models using climate variables as predictors. BMC Infect Dis 2011,
development, preliminary validation, and samples of simulation results. 11:166.
Am J Trop Med Hyg 1995, 53:489–506. 37. Thammapalo S, Chongsuwiwatwong V, McNeil D, Geater A: The climatic
14. Kuno G: Review of the factors modulating dengue transmission. factors influencing the occurrence of dengue hemorrhagic fever in
Epidemiol Rev 1995, 17:321–335. Thailand. Southeast Asian J Trop Med Public Health 2005, 36:191–196.
15. McMichael AJ, Haines A: Global climate change: the potential effects on 38. Patz JA, Martens WJ, Focks DA, Jetten TH: Dengue fever epidemic
health. Br Med J 1997, 315:805–809. potential as projected by general circulation models of global climate
16. Patz JA, Epstein PR, Burke TA, Balbus JM: Global climate change and change. Environ Health Perspect 1998, 106:147–153.
emerging infectious diseases. JAMA 1996, 275:217–223. 39. Chaves LF, Koenraadt CJM: Climate change and highland malaria: fresh air
17. Hii YL, Rocklöv J, Ng N, Tang CS, Pang FY, Sauerborn R: Climate variability for a hot debate. Q Rev Biol 2010, 85:27–55.
and increase in intensity and magnitude of dengue incidence in 40. Cazelles B, Chavez M, McMichael AJ, Hales S: Nonstationary influence of El
Singapore. Glob Health Action 2009, 11:2. doi: 10.3402/gha.v2i0.2036. Nino on the synchronous dengue epidemics in Thailand. PLoS Med 2005,
18. Sarfraz MS, Tripathi NK, Tipdecho T, Thongbu T, Kerdthong P, Souris M: 2:e106.
Analyzing the spatio-temporal relationship between dengue vector 41. Colon-Gonzalez FJ, Lake IR, Bentham G: Climate variability and dengue
larval density and land-use using factor analysis and spatial ring fever in warm and humid Mexico. Am J Trop Med Hyg 2011, 84:757–763.
mapping. BMC Public Health 2012, 12:1–19. 42. Hu W, Clements A, Williams G, Tong S: Dengue fever and El Nino/
19. Aziz AL, Dieng H, Ahmad AH, Mahyoub JM, Turkistani AM, Mesed H, Koshike S, Southern Oscillation in Queensland, Australia: a time series predictive
Satho T, Salmah MRC, Ahmad H, Zuharah WF, Ramli AS, Miake F: Household model. Occup Environ Med 2010, 67:307–311.
survey of container-breeding mosquitoes and climatic factors influencing 43. Hu W, Clements A, Williams G, Tong S, Mengersen K: Spatial patterns and
the prevalence of Aedes aegypti (Diptera: Culicidae) in Makkah City, Saudi socioecological drivers of dengue fever transmission in Queensland,
Arabia. Asian Pac J Trop Biomed 2012, 2:849–857. Australia. Environ Health Perspect 2012, 120:260–266.
20. Gubler DJ: Dengue and West Nile virus–an interview with Duane Gubler, 44. Bambrick HJ, Woodruff RE, Hanigan IC: Climate change could threaten
Sc.D., reported by Vicki Glaser. Vector Borne Zoonotic Dis 2001, 1:81–88. blood supply by altering the distribution of vector-borne disease:
21. Scott TW, Morrison AC, Lorenz LH, Clark GG, Strickman D, Kittayapong P: an Australian case-study. Glob Health Action 2009, 2. doi: 10.3402/gha.
Longitudinal studies of Aedes aegypti (Diptera: Culicidae) in Thailand and v2i0.2059.
Puerto Rico: population dynamics. J Med Entomol 2000, 37:77–88. 45. Chowell G, Cazelles B, Broutin H, Munayco CV: The influence of geographic
22. Chaves LF, Scott TW, Morrison AC, Takada T: Hot temperatures can force and climate factors on the timing of dengue epidemics in Peru, 1994–2008.
delayed mosquito outbreaks via sequential changes in Aedes aegypti BMC Infect Dis 2011, 11:164.
demographic parameters in autocorrelated environments. Acta Trop 46. Bulto PL, Rodriguez AP, Valencia AR, Vega NL, Gonzalez MD, Carrera AP:
2014, 129:15–24. Assessment of human health vulnerability to climate variability and
23. Carrington LB, Seifert SN, Willits NH, Lambrechts L, Scott TW: Large diurnal change in Cuba. Environ Health Perspect 2006, 114:1942–1949.
temperature fluctuations negatively influence Aedes aegypti (Diptera: 47. Hales S, Weinstein P, Souares Y, Woodward A: El Nino and the dynamics
Culicidae) life-history traits. J Med Entomol 2013, 50:43–51. of vectorborne disease transmission. Environ Health Perspect 1999,
24. Carrington LB, Seifert SN, Arrmijos MV, Lambrechts L, Scott TW: Reduction 107:99–102.
of Aedes aegypti vector competence for dengue virus under large 48. Johansson MA, Dominici F, Glass GE: Local and global effects of climate
temperature fluctuations. Am J Trop Med Hyg 2013, 88:689–697. on dengue transmission in Puerto Rico. PLoS Negl Trop Dis 2009, 3:e382.
25. Carrington LB, Seifert SN, Arrmijos MV, Lambrechts L, Scott TW: Fluctuations 49. Chen M-J, Lin C-Y, Wu Y-T, Wu P-C, Lung S-C, Su H-J: Effects of extreme
at a low mean temperature accelerate dengue virus transmission by precipitation to the distribution of infectious diseases in Taiwan,
Aedes aegypti. PLoS Negl Trop Dis 2013, 7:e2190. 1994–2008. PLoS One 2012, 7:e34651–e34651.You can also read
