Climate Variability, Social and Environmental Factors, and Ross River Virus Transmission: Research Development and Future Research Needs

Background Arbovirus diseases have emerged as a global public health concern. However, the impact of climatic, social, and environmental variability on the transmission of arbovirus diseases remains to be determined. Objective Our goal for this study was to provide an overview of research development and future research directions about the interrelationship between climate variability, social and environmental factors, and the transmission of Ross River virus (RRV), the most common and widespread arbovirus disease in Australia. Methods We conducted a systematic literature search on climatic, social, and environmental factors and RRV disease. Potentially relevant studies were identified from a series of electronic searches. Results The body of evidence revealed that the transmission cycles of RRV disease appear to be sensitive to climate and tidal variability. Rainfall, temperature, and high tides were among major determinants of the transmission of RRV disease at the macro level. However, the nature and magnitude of the interrelationship between climate variability, mosquito density, and the transmission of RRV disease varied with geographic area and socioenvironmental condition. Projected anthropogenic global climatic change may result in an increase in RRV infections, and the key determinants of RRV transmission we have identified here may be useful in the development of an early warning system. Conclusions The analysis indicates that there is a complex relationship between climate variability, social and environmental factors, and RRV transmission. Different strategies may be needed for the control and prevention of RRV disease at different levels. These research findings could be used as an additional tool to support decision making in disease control/surveillance and risk management.

The global climate is changing at an unprecedented rate, with rising surface temperatures, melting glaciers, rising sea levels, and changes in climate variability and extreme events. Questions about its possible public health consequences command increasing attention. It has become evident that global climate change is altering-indeed, will continue to alter-the pattern of both average climatic conditions and variability, although the projected trends within specific geographic regions remain uncertain (Intergovernmental Panel on Climate Change 2007; Royer et al. 2007). There are unresolved questions about a) how climate change may affect the transmission of vector-borne diseases, and b) to what extent the empirical evidence of shortterm climate variation and disease transmission can be applied to the estimation of future impacts of climate change.
Arboviral infections are a global health problem accounting for significant morbidity and mortality in human populations (Ansari and Shope 1994;Gubler 2002). Ross River virus (RRV) infection is the most common and widespread arboviral disease in Australia and some Pacific island nations, with thousands of clinical cases occurring each year [Australian Department of Health and Ageing (ADHA) 2008; Mackenzie et al. 1998]. Over the last 15 years (i.e., 1993-2007), 64,562 laboratoryconfirmed RRV cases (4,304/year) have been reported in Australia alone (ADHA 2008). Of these, more than half were reported in Queensland, a subtropical and tropical state. In recent decades, RRV outbreaks have increased, notably in urban estates and tourist destinations (Harley et al. 2001;Kelly-Hope et al. 2004a;. For example, in Australia, there were 8,387 notifications of mosquitoborne diseases (MBDs) during the period 1 July 2005 to 30 June 2006, which was the second highest on record since 1995-1996 (Liu et al. 2006). RRV infections accounted for about 70% of MBD notifications with an onset in [2005][2006]. The major behavioral, environmental, and ecologic risk factors for RRV disease transmission include outdoor activities (e.g., camping), climate variability (e.g., increasing rainfall and temperature), people moving from nonendemic into endemic areas, and the effectiveness of vector control programs. The reasons for the increased RRV activity may include urban developments in or near wetland and salt marsh habitats, increased travel, and socioecologic changes (Muhar et al. 2000;Russell 2002;Tong et al. 2004).
Because RRV infection is the most common and widespread mosquito-borne disease in Australia, there has been an increasing research interest in the assessment of major determinants of RRV transmission (Hu et al. , 2006a(Hu et al. , 2006bJacups et al. 2008, Kelly-Hope et al. 2004aTong et al. , 2005Woodruff et al. 2006). To identify current knowledge gaps and research needs, we critically reviewed the impact of climatic, social, and environmental variability on the transmission of RRV disease. Here we provide an overview of research development and future research directions in this emerging field. burden to infected residents, particularly poor people (Ratnayake 2005). Outbreaks of RRV impact considerably on tourism and industry, as well as on communities (Harley et al. 2001;Hawkes et al. 1985;Mackenzie et al. 1998). Additionally, there is substantial expenditure for vector control programs each year (e.g., > $10 million in Queensland in 2004 alone) (Tomerini 2007). Further, there is an increasing concern about the risk of introducing RRV to neighboring countries such as New Zealand (Kelly-Hope et al. 2002) and the global emergence/resurgence of arbovirus diseases (Gubler 2002). Therefore, it is important to improve the public health strategy to control and prevent this widespread disease in the Oceania region (Guillaumot 2005;McMichael et al. 2003).
The ecology of RRV is complex. The virus and its reservoir host species, the vector, the human population, and environmental conditions play key roles in the RRV transmission cycles. RRV is an alphavirus that was named after it was first isolated from Ocherotatus vigilax mosquitoes collected around Ross River near Townsville in northern Queensland in 1959 (Doherty et al. 1963). The natural vertebrate hosts for this disease include marsupial animals (e.g., kangaroos, wallabies) and possibly other animals (e.g., dogs, cats, horses, possums). The incubation period may be as long as 21 days or as short as 3 days (usually 7-9 days) (Harley et al. 2002). More than 30 mosquito species have been implicated as vectors of RRV (Mackenzie et al. 1994;Russell 1994). Ocherotatus vigilax and Ocherotatus camptorhynchus, which breed in intertidal wetlands, are important in coastal regions, as are floodwater Aedes species in many inland areas. Culex annulirostris, which breeds in vegetated semipermanent and permanent fresh water, is common in areas of the tropics and temperate regions that are subject to flooding or irrigation during summer. Species such as Ocherotatus notoscriptus may be important in semirural and urban areas (Harley et al. 2002;Mackenzie et al. 1994;Russell 1994).
The virus is dependent on the continuing recruitment of nonimmune hosts in the reservoir population. The distribution and abundance of the reservoir population will affect the availability of viremic individuals to mosquitoes, and a nonimmune reservoir population leads to increased virus activity (Weinstein 1997). A number of mosquitorelated factors (e.g., mosquito species, mosquito abundance) also influence the level of RRV activity. The human population is susceptible to RRV infection if individuals are nonimmune and are exposed to the virus at the reservoir/mosquito/human interface. Such exposure is enhanced by human intrusions into natural ecosystems by the expansion of agriculture, forestry, tourism, and building of new suburbs in or near wetlands (Russell 2002;Weinstein 1997).
Weather conditions directly affect the breeding, survival, and abundance of mosquitoes and the extrinsic incubation period of the disease. In seasons with high temperatures and rainfall, the vegetation upon which intermediate hosts such as kangaroos, possums, and horses depend will flourish, and more nonimmune reservoir hosts will be added to the temporally and spatially expanding population (Harley et al. 2001;Hu et al. 2004;Kelly-Hope et al. 2004a;Russell 2002). RRV disease has strong spatial and temporal patterns, because mosquito density and longevity depend on a number of environmental and social factors (e.g., temperature, precipitation, mosquito-breeding habitats, vector control programs) (Gatton et al. 2005;Hu et al. 2005;. RRV is considered to be one of the few infectious diseases that can be predicted by climate-based early warning systems (EWS) [World Health Organization (WHO) 2004], although predictions may vary because of local conditions (Mackenzie et al. 2000). Climate variability strongly affects the replication of the virus; the breeding, abundance, and survival of the mosquito species; the breeding patterns of major hosts; and people's behavior (Harley et al. 2001;Tong et al. 2004). These variables can be modeled to predict the onset and severity of disease epidemics at different temporal and spatial scales (Hightower et al. 1998;Moore and Carpenter 1999). However, in developing EWS for MBDs and other infectious diseases, it is important to also include nonclimatic influences in the models and to make the models relevant to particular response decisions and to the needs of policy makers (WHO 2004). Below we discuss exemplary research on the complex interactions relevant to developing an EWS ( Figure 1).

Climate Variability, Social and Environmental Factors, and RRV Transmission
Identifying relevant studies. To identify all relevant studies (in addition to our own research), we conducted a systematic literature search on climatic, social, and environmental factors and RRV disease. Potentially relevant studies were identified from a series of electronic searches. Databases searched were MEDLINE (via EBSCOhost; EBSCO Publishing, Ipswich, MA), Current Contents Connect (via ISI Web of Knowledge; Thomson Reuters, London, UK), and ScienceDirect (Elsevier B.V., Amsterdam, the Netherlands). Literature searches were conducted in November 2007, with no date or language restriction. The databases were searched with the following key words or MeSH terms: "Ross River virus" AND "climate" OR "social" OR "environmental" OR "climatic" OR "rainfall" OR "temperature." In addition, in June 2008 we performed a further complementary search of articles published since 2007. Titles and abstracts for all the publications identified by the literature search were first assessed for suitability for inclusion. Full copies of all suitable studies were then obtained and critically reviewed.

Key predictors of RRV transmission: integration of our own research with literature.
The primary objectives of our own research were to assess the impact of climate variability and social and environmental factors on the transmission of RRV disease, and their role in developing EWS for the control and prevention of this most common and widespread MBD in Australia. We also identified several other studies on this topic through the literature search (Table 1). We collected a large amount of data from relevant government agencies on climate variables, social and environmental factors, and notified cases of RRV disease for Queensland, where most cases are usually reported ( Figure 2). A range of geographic information system (GIS) techniques and ecologic timeseries models were performed to assess the impact of climate variability on RRV transmission in Queensland (Hu et al. , 2006a(Hu et al. , 2006b(Hu et al. , 2007Tong et al. 1998Tong et al. , 2004Tong et al. , 2005Tong and Hu 2002).
In a previous study, we used GIS techniques to assess the distribution of RRV in Queensland . For that study, we obtained the computerized data set on the notified RRV cases in Queensland for the period of 1985-1996 and key sociodemographic  (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995) correlation in minimum temperature), rainfall, included humidity, and SOI were positively associated with the incidence of RRV disease Tong and Hu 2001 Cairns, Queensland Time-series analysis ARIMA model Rainfall and relative humidity appeared Only one coastal city was (1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996) to play significant roles in the included transmission of RRV disease in Cairns Queensland (1985-1996 Time-series analysis ARIMA model Overall, rainfall, temperature, and tidal The magnitude of the levels were important environmental climate-RRV relationship determinants in the transmission cycles varied across the eight of RRV disease across Queensland major cities in Queensland Done et al. 2002Queensland (1991-1997 Time-series analysis Correlograms and The quasi-biennial cycle accounts for Spatial variation was not periodograms 77% of the variance in RRV cases. examined Tong and Hu 2002 Coastline and inland Time-series analysis Poisson regression Maximum temperature exhibited The relation between climate regions in Queensland a greater impact on the RRV variables and RRV needs to (1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996) transmission in coastline than in be viewed within a wider inland cities; minimum temperature context of other socioand relative humidity seemed to environmental variability affect the RRV transmission more at the inland than the coastline. Woodruff et al. 2002 Southeastern Australia Ecologic study Logistic regression Early warning of weather conditions The sensitivity of the model (1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999) conducive to outbreaks of RRV varied with area disease is possible Tong et al. 2004 Townsville region Time-series analysis SARIMA model Rainfall, high tide, and maximum Spatial variation was not (1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996) temperature were likely to be key examined determinants of RRV transmission in the Townsville region Hu et al. 2004Brisbane (1985 Time-series analysis SARIMA model Monthly precipitation was significantly This is a broad, ecologic associated with RRV transmission but assessment at the city had no significant association for other level; more detailed risk climate variables (e.g., temperature, assessment at community relative humidity, high tides) and individual levels may also be required Kelly-Hope et al.
Four geoclimatic regions Ecologic study Descriptive statistics Rainfall in outbreak years tended to be It is unclear how RRV 2004a in Australia (various above average and higher than rainfall outbreak years were periods from 1971 in nonoutbreak years. Overall defined across four regions to 1997) temperatures were warmer during outbreak years; however, seasonal and monthly trends differed across geoclimatic regions of the country Kelly-Hope et al.
Australia  Analysis of historical Descriptive statistics The magnitude, regularity, seasonality, Information bias is likely to 2004b reports and locality of outbreaks ranged widely; occur on how to define and environmental conditions act differently record outbreaks for such a in tropical, arid, and temperate regions. long period Overall, rainfall seems to be the single most important risk factor. Tong et al. 2005Brisbane (1998 Time-series analysis Poisson regression There were complex interrelationships Only one metropolitan city between rainfall, mosquito density, and was included RRV transmission Gatton et al. 2005Queensland (1991 Ecologic study Logistic regression The variables identified as important in Selection bias might occur in predicting RRV disease outbreaks choosing different stations differed between regions and also for different local between summer and autumn government areas Hu et al. 2006aBrisbane (1998 Time-series analysis Polynomial Both rainfall and mosquito density were Only one metropolitan city distributed lag and strong predictors of the RRV was included. SARIMA models transmission Woodruff et al. 2006 Western Australia Ecologic study Logistic regression Mosquito surveillance data could increase Only a temperate region was (1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999) the accuracy of disease prediction models included Jacups et al. 2008 Darwin region (1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006) Time-series analysis Poisson regression The best global model included rainfall, Only a tropical region was minimum temperature, and three included mosquito species and can accurately predict RRV infections throughout the year in the Darwin region Abbreviations: ARIMA, auto-regressive integrated moving average; SARIMA, seasonal auto-regressive integrated moving average; SOI, Southern Oscillation Index. a By chronological order of publication.
information from Queensland Health (Brisbane, Queensland, Australia). The reported place of onset for each case was used to characterize the geographic distribution of the RRV infection within Queensland. We used MapInfo Professional (Ersis Australia, Sydney, Australia) to display the spatial and temporal distributions of RRV cases. The digital base map data sets used for constructing the GIS were obtained primarily from the Queensland Department of Natural Resources (Brisbane, Queensland, Australia). The onset places in the data set were geocoded to the digital base maps of localities using MapInfo and Microsoft Access (Microsoft Corporation, Redmond, WA, USA) software. The location for each notified case of the RRV infection was then obtained by overlaying the database of the onset places of notified RRV infections with the digital base maps. In that study ), we found RRV notifications from 489 localities between 1985 and 1988. By the period 1989-1992, this had increased by 65% to 805 localities and, by 1993-1996, it had increased to 1,157 localities, an increase of 137% from the 1985-1988 period (Figure 3). The geographic distribution of the notified RRV cases has apparently expanded in Queensland over recent years, as shown in Figure 3, and it is unlikely to be entirely explained by better notification and laboratory reporting, because the geographic distribution of RRV varied substantially in Queensland. That study ) was the first to use GIS techniques to display the variation of the detailed spatiotemporal distribution of RRV disease. The study suggested that the geographic expansion of RRV may be associated with many factors, including population growth, urban sprawl, increased travel, and socioecologic change.
In another study , we performed an ecologic time-series analysis to examine the association between climate variability and the transmission of RRV disease between 1985 and 1996 in Queensland. Information on the notified RRV cases was obtained from the Queensland Health, and climate and population data were supplied by the Australian Bureau of Meteorology and the Australian Bureau of Statistics, respectively. The autoregressive integrated moving average model was used to examine the relation between climate variability, tides, and the monthly incidence of notified RRV infections. As maximum and minimum temperatures were highly correlated with each other (r s = 0.75), two separate models were developed. For the eight major cities in Queensland, the climate-RRV correlation coefficients were in the range of 0.12-0.52 for maximum and minimum temperatures, -0.10 to 0.46 for rainfall, and 0.11-0.52 for relative humidity and high tide . For the whole state, rainfall [partial regression coefficient, 0.017 (95% confidence interval, 0.009-0.025) in Model I and 0.018 (0.010-0.026) in Model II], and high tidal level [0.030 (0.006-0.054) in Model I and 0.029 (0.005-0.053) in Model II] seemed to have played significant parts in the transmission of RRV in Queensland. Maximum temperature was also marginally significantly associated with the incidence of RRV infection. The results of our research  showed that although many factors can affect the RRV transmission cycles, RRV disease is generally sensitive to climate variability. At a macro level, rainfall, temperature, and tidal levels appeared to be important environmental determinants in the transmission cycles of RRV disease. The findings are generally consistent with other studies (Hu et al. , 2006a(Hu et al. , 2007Kelly-Hope et al. 2004a;Mackenzie et al. 2000;Russell 2002;Tong et al. , 2004Tong et al. , 2005Tong and Hu 2002). This information may be useful for designing and implementing RRV interventions at the state and national levels.
However, the magnitude and nature of the association between climate variables and social and environmental factors and RRV transmission seemed to vary with geographic area (Kelly-Hope et al. 2004b;Tong and Hu 2002). For example, we assessed the difference in the potential predictors of the RRV incidence in coastline and inland regions of Queensland using time-series regression models (Tong and Hu 2002). The function of cross-correlations was used to compute a series of correlations between climate variables (rainfall, maximum temperature, minimum temperature, relative humidity, and high tide) and the monthly incidence of RRV disease over a range of time lags. In that study (Tong and Hu 2002), time-series regression models were performed to adjust for the autocorrelations of the monthly incidences of RRV disease and the confounding effects of seasonality, the case notification time, and population sizes. The results showed that the incidence of RRV disease was significantly associated with rainfall, maximum temperature, minimum temperature, relative humidity, and high tide in the coastline region, and with rainfall and relative humidity in the inland region. There was a significant interaction between climate variables and locality in RRV transmission. Different responses of RRV to climate variability were observed between coastal and inland cities in Queensland (Tong and Hu 2002). Overall, rainfall appeared to be the single most important determinant of RRV transmission. However, maximum temperature appeared to exhibit greater impacts on the RRV transmission in coastal cities than in inland cities. Minimum temperature and relative humidity seemed to affect the RRV transmission inland more than at the coast.
Kelly-Hope et al. (2004b) also compared seasonal and monthly rainfall and temperature trends in outbreak and nonoutbreak years at four epidemic-prone locations in Australia. Their analyses showed that rainfall in outbreak years tended to be above average and higher than rainfall in nonoutbreak years. Overall temperatures were warmer during outbreak years. However, there were a number of distinct deviations in temperature, which seemed to play a role in either promoting or inhibiting outbreaks (Kelly-Hope et al. 2004b). These Tong et al. 1594 VOLUME 116 | NUMBER 12 | December 2008 • Environmental Health Perspectives   [1993][1994][1995][1996] results showed that climatic differences occur between outbreak and nonoutbreak years; however, seasonal and monthly trends differed across geoclimatic regions of the country. Now, considerable evidence has accrued to show that precipitation is an important factor in the transmission of RRV in many parts of Australia. All mosquitoes have aquatic larval and pupae stages and therefore require water for breeding (Hawkes et al. 1985;Muhar et al. 2000). It is precipitation that determines the presence or absence of breeding sites. Rainfall events and subsequent floods can lead to outbreaks of arboviral disease, largely by enabling breeding of vector mosquitoes (Hawkes et al. 1985). In general, epidemic activity of arbovirus is more often observed in temperate areas with heavy rainfall, flooding, or high tides, whereas in tropical Australia transmission occurs throughout the year (Russell 2002).
A number of studies have attempted to develop predictive models for the transmission of RRV disease using routinely collected data on climate and tidal variables and mosquito and disease surveillance (Hu et al. 2006a;Tong et al. 2005;Woodruff et al. 2006). Our recent research at the city level reveals that in some areas (e.g., the Brisbane metropolitan area), rainfall and sea tides directly influence the density of mosquitoes such as Culex annulirostris and Aedes vigilax, respectively, which then affects the transmission patterns of RRV in inland and coastal suburbs (Hu et al. 2006a;Tong et al. 2005). For Brisbane City, we obtained monthly data on the counts of RRV cases, monthly total rainfall, human population size, and mosquito density (i.e., average number of mosquitoes trapped in all mosquitomonitoring stations per month) between 1 November 1998 and 31 December 2001 from the Queensland Department of Health, Australian Bureau of Meteorology, Australia Bureau of Statistics, and Brisbane City Council, respectively. Both polynomial distributed lag time-series regression and seasonal auto-regressive integrated moving average (SARIMA) models were used to examine associations of RRV transmission with rainfall and mosquito density after adjustment for seasonality and auto-correlation. The results show that 85% and 95% of the variance in the RRV transmission was accounted for by rainfall and mosquito density, respectively (Hu et al. 2006a). Both rainfall and mosquito density were strong predictors of the RRV transmission in simple models. However, multivariate polynomial distributed lag models show that only mosquito density at lags of 0 and 1 month was significantly associated with the transmission of RRV disease. The SARIMA models show similar results (Figure 4) (Hu et al. 2006a). Woodruff et al. (2006) focused on a temperate region of western Australia between July 1991 and June 1999. Both early and later warning logistic regression models were developed to test the sensitivity of data on environment (tide height, rainfall, sea surface temperature) and mosquito counts for predicting epidemics of disease. Environment data alone were moderately sensitive (64%) for predicting epidemics during the early warning period. Addition of mosquito surveillance data increased the sensitivity of the early warning model to 90%. The later warning model had a sensitivity of 85%. Woodruff et al. (2006) concluded that the environment data they used are relatively cheap to collect and useful for the prediction of RRV disease epidemics. Mosquito surveillance data provide a more expensive early warning but add substantial predictive value.
Recently, Jacups et al. (2008) described the epidemiology of RRV infection in the endemic Darwin region of tropical northern Australia and developed a predictive model for RRV infections. Laboratory-confirmed cases of RRV infection between 1 January 1991 and 30 June 2006 were analyzed, together with climate, tidal, and mosquito data collected weekly over the study period from 11 trap sites around Darwin, using both correlations and Poisson regression models. Correlations revealed strong associations between monthly RRV infections and climatic variables and also each of the four implicated mosquito species populations. The best global model included rainfall, minimum temperature, and three mosquito species, which explained 63.5% deviance and predicted disease accurately. The results also indicate that predicted anthropogenic global climatic changes may result in an increase in RRV infections (Jacups et al. 2008).
These research findings provide further evidence that mosquito density, largely influenced by rainfall and high tide, is an important risk factor for RRV transmission. Increased mosquito density increases the likelihood that a person will be bitten and, hence, this affects the risk of contracting a mosquito-borne disease. Improved understanding of the relationship between environmental variability, mosquito density, and RRV transmission might assist disease control managers in planning and implementing public health interventions. Results of studies such as this might facilitate the development of EWS for reducing the incidence of this widespread disease and other MBDs in Australia and other Pacific island nations.

Other Risk Factors
Individual and behavioral risk factors. Harley et al. (2005) examined individual risk and protective factors for RRV disease in a highincidence area. They performed a prospective matched case-control study with new community cases of RRV disease in the local government areas of Cairns, Mareeba, Douglas, and Atherton, in tropical Queensland, from 1 January to 31 May 1998. Protective measures against mosquitoes reduced the risk for disease. Mosquito coils, repellents, and citronella candles each decreased risk by at least 2-fold, with a dose response for the number of protective measures used. Light-colored clothing decreased risk 3-fold, and camping increased the risk 8-fold (Harley et al. 2005). These risks were substantial and statistically significant, and provide a basis for educational programs on individual protection against RRV disease in Australia.
Genetic divergence. Sammels et al. (1995) conducted a molecular epidemiologic study to examine the evolution of RRV in Australia and the Pacific Islands. Nucleotide sequences of the E2 and E3 genes of five RRV strains revealed remarkable conservation between  Year Mosquito density/rainfall (mm) RRV incidence rate (1/100,000)

RRV incidence Rainfall
Mosquito density of only 3.3%. Sequence data from a 505-base pair fragment of the E2 gene from 51 additional strains showed that RRV has diverged genetically into three separate groups, although at least 95% sequence homology was still maintained between all 56 strains. Each genetic type predominates in a particular geographic region of Australia and can be broadly defined as occurring in the western, northeastern, and southeastern regions of Australia. However, some RRV strains did not follow this pattern of geographic distribution, indicating movement of virus by the travel of viremic humans or livestock across the continent. The Pacific Islands isolates all belong to the southeastern genotype. The findings of Sammels et al. (1995) suggest genetic divergence and independent evolution of RRV within geographically isolated enzootic foci; however, selective pressures maintain high nucleotide conservation in nature.

Vector Control Programs
As outbreaks of RRV impact considerably on tourism and industry, as well as on communities (Harley et al. 2001;Russell 2002), there is substantial expenditure for vector control programs each year (e.g., > $10 million in Queensland in 2004 alone) (Tomerini 2007). Tomerini also showed that the costs of mosquito control may be offset by reduced incidence (and cost) of RRV disease. There is a need to examine how to undertake vector control programs more effectively and efficiently. For example, more resources should be mobilized into high-risk areas during highrisk periods but fewer resources are needed in low-risk areas or during nonepidemic periods. At present, a similar amount of funding is provided to implement vector control activity in most areas each year regardless of epidemic or nonepidemic circumstances.

Development of EWS
EWS have been developed and applied in many fields, including food, agriculture, and natural disasters (e.g., tsunamis). The purpose of EWS is to reduce vulnerability and increase preparedness (WHO 2004). Over recent years, model-driven EWS have been attempted increasingly in epidemiology and public health, particularly for the control and prevention of infectious diseases. In general, EWS are regarded as management tools to predict outbreaks of infectious diseases so that health authorities can prepare for the epidemics and act accordingly to mitigate against or avoid the impact of these diseases (Teklehaimanot et al. 2004a(Teklehaimanot et al. , 2004b. The development of spatiotemporal models has been rapidly evolving in the area of control and prevention of MBDs. Some models have been successfully developed to predict the likelihood of MBD epidemics using weather and environmental data (Teklehaimanot et al. 2004a(Teklehaimanot et al. , 2004bThomson et al. 2006). RRV disease has strong spatial and temporal patterns, because mosquito density and longevity depend on a number of environmental and social factors (e.g., temperature, precipitation, mosquito-breeding habitats, vector control programs) Kelly-Hope et al. 2004a;. The pattern of RRV and other MBDs is likely to change with the changing socioecologic conditions (e.g., urbanization, increasing travel, global climate change) (Russell 1998;Woodruff et al. 2006). Most published reviews indicate that RRV may be influenced by climate change, and the impact may vary from region to region (Harley et al. 2001;Russell 2002). However, although RRV is the most common and widespread MBD in Australia and some Pacific island nations, little research has been conducted to develop an integrated EWS to forecast the likelihood of RRV outbreaks in different socioecologic regions, particularly in new urban estates and popular tourist areas.
Thus, there is a need to facilitate shortterm epidemic forecasting and to improve scenario-based predictive modeling for the control and prevention of RRV and other MBDs to enhance biosecurity, to better adapt to rapid socioecologic changes, and to minimize the adverse public health impact of these changes. Climate variability strongly affects the replication of the virus, the breeding, abundance, and survival of the mosquito species, the breeding patterns of major hosts, and human behavior (Harley et al. 2001;Russell 1998;Tong 2004). These variables can be modeled to predict the onset and severity of disease epidemics at different temporal and spatial scales. Further, it is anticipated that the properly developed EWS for RRV disease transmission will improve our understanding of biologic/ecologic mechanisms of disease outbreaks and may have wide applications in planning RRV and other disease control and risk management programs.
Disease control measures such as EWS have great public health implications. Comprehensive and interdisciplinary EWS can greatly assist in improving vector control and personal protection to reduce the likelihood of RRV epidemics, as the previous models generally provided 1-3 months of advance warning (Hu et al. , 2006a(Hu et al. , 2007Tong et al. 1998Tong et al. , 2004Tong et al. , 2005Tong and Hu 2002). For example, differentially increasing insecticide spraying in high-risk areas during high-risk periods and decreasing it in low-risk areas and periods will improve cost effectiveness of vector control operations. If anticipating an outbreak of RRV disease, decision makers in disease control programs (e.g., communicable disease managers) can increase vigilance (e.g., by alerting district health offices), enhance vector control activity, request more frequent reporting to facilitate early identification of any problem, and initiate community education programs promptly in the affected areas. Additionally, the novel methods developed in these studies are also potentially useful for other MBDs (e.g., malaria, dengue fever, Japanese encephalitis) and may assist health authorities in determining public health priorities more wisely and using resources more efficiently, especially in the context of climate and sea level changes. Although EWS have significant implications in disease control and surveillance, such measures have not been formally implemented in Australia. There is an urgent need to bridge the gaps between science and policy.

Conclusions
A growing body of evidence suggests that, even though many factors can affect the transmission cycles of RRV disease, climate and tidal variables (including rainfall, temperature, and tidal levels) and mosquito density are generally important environmental predictors of RRV disease. However, the nature and magnitude of such associations may vary with geographic area and socioenvironmental condition, and thus, further in-depth research on socioenvironmental change and RRV transmission is required at local, regional, and national levels.
An assessment of factors predicting RRV disease transmission will help local authorities identify periods of high risk, optimizing the provision of additional mosquito control measures and community education. Climate data are relatively cheaper and easier to collect than other data (e.g., mosquito density and species). It is possible to improve the effectiveness of public health responses through the prediction of RRV epidemics with the appropriate integration and analysis of weather forecast, social and environmental factors, and disease surveillance data. However, the relationship between climate variables and RRV needs to be viewed within a wider context of other social and environmental changes. To advance this field, key research priorities include • Methodologic development for quantifying the relative importance of climate variability/change and other social and environmental factors in outbreaks of RRV disease • Further in-depth research on socioenvironmental change and RRV transmission at different levels, taking into account the nature and magnitude of the relationship between climate variability and RRV transmission that may vary with geographic area and socioenvironmental condition (it is important to integrate both climatic and nonclimatic factors in the development of EWS) • Development of scenario-based risk assessment models to project possible impacts of climate change on RRV transmission from local to national • Develop different public health intervention strategies to control and prevent this disease at different levels to address RRV epidemiologic patterns throughout Australia (it is important to evaluate what public health intervention strategies are effective to control and prevent this disease at different levels). The prospect of a changing regional climate is likely to affect the seasonal and geographic distribution of RRV. Projected anthropogenic global climatic change may result in an increase in RRV infections. The details of these changes will be difficult to predict at a regional level, but the development of EWS should allow actions to reduce the incidence of the disease, offsetting any increases arising from the changing climate. Thus, the development of these systems can help us cope with RRV in the current climate and help us adapt to the consequences of climate change in the future, at minimal cost. Of course, care will need to be taken to ensure that climate changes do not, of themselves, change the underlying climate/heath/societal relationships that form the basis of the EWS.
Clearly, a concerted, interdisciplinary, multisectoral effort is required to design and implement these research activities. We anticipate that climate/RRV research will provide an inspirational opportunity for researchers, policy makers, and other stakeholders to work together more closely to control and prevent this widespread disease.