Portal de Periódicos da CAPES

Abundance of West Nile virus mosquito vectors in relation to climate and landscape variables

2011; Wiley; Volume: 36; Issue: 1 Linguagem: Inglês

10.1111/j.1948-7134.2011.00143.x

1948-7134

Jayne M. Deichmeister, Aparna Telang,

Malaria Research and Control

It is currently unclear if the potential for West Nile virus transmission by mosquito vectors in the eastern United States is related to landscape or climate factors or both. We compared abundance of vector species between urban and suburban neighborhoods of Henrico County, VA, in relation to the following factors: temperature, precipitation, canopy cover, building footprint, and proximity to drainage infrastructure. Mosquitoes were collected throughout the 2005, 2006, and 2007 seasons and tested for West Nile virus (WNV) in pools of 10–50. Test results of mosquito pools were compared to average site abundance from 37 sites in Henrico County, VA; abundance was then examined in relation to ecological variables. Urban infrastructure was positively correlated with the abundance of Culex pipiens L./Cx. restuans, and our findings implicate combined sewer overflow systems as large contributors to Culex vector populations. No measure of urbanization examined in our study was correlated with Aedes albopictus abundance. Our study showed that certain landscape variables identified using Geographic Information Systems are valuable for predicting primary WNV vector abundance in Virginia, and that temperature along with low precipitation are strong predictors of population growth. Our results support other regional studies that found WNV proliferates under drought conditions. West Nile virus (WNV), a mosquito-borne flavivirus that is indigenous to Asia, Africa, Europe, and Australia, has historically been ignored as a significant human pathogen because most cases display mild or no symptoms (Hayes 2001). An outbreak in the northeastern United States in 1999, however, was the first occurrence of the disease in the western hemisphere (Lanciotti et al. 1999) and involved hundreds of cases of encephalitis and meningoencephalitis in urban areas (Hayes 2001). Since the 1999 outbreak, WNV has spread throughout all of the eastern United States and parts of the west (Campbell et al. 2002) and has caused counties in almost every state to invest in municipal surveillance programs to control the mosquito populations that transmit this pathogen. Surveillance programs attempt to monitor the distribution of vector species and reduce their populations by treating breeding sites with species-targeted larvicides. West Nile virus is primarily maintained in nature by transmission cycles between ornithophilic mosquitoes and passerine avian hosts. WNV can occasionally infect and cause disease in other vertebrate hosts, including humans and horses, but such hosts are considered tangential because they do not generate sufficient viremia to infect mosquito hosts. The tangential cycle usually involves other mosquito species that will take blood from a variety of animal hosts, thereby posing a considerable threat to humans by acting as bridge vectors (Kilpatrick et al. 2005). Enzootic outbreaks significantly increase the likelihood of human infection by increasing the odds that a bridge vector will feed on an infected bird. Public health authorities have singled out Culex pipiens L. (Diptera: Culicidae) and Culex restuans Theobald (Diptera: Culicidae) as main vectors because not only are they ornithophilic, but they are competent pathogen hosts, frequently infected in the natural environment, and have established native populations where human infection has repeatedly occurred (Spielman 2001, Kilpatrick et al. 2005). Culex larval habitats remain one of the largest targets for municipal surveillance programs in the United States, but local drainage infrastructure greatly increases the proliferation of suitable sites. The use of open roadside ditches is one example; Culex pipiens seem to prefer transient pools rich in organic matter (Jackson et al. 2005) thus rendering these ditches as attractive oviposition sites (White 2001). While Culex species have been implicated as primary enzootic vectors, the Asian tiger mosquito, Aedes albopictus Skuse (Diptera: Culicidae), is described as a highly competent vector of WNV (Turrell et al. 2001) and its diverse host preferences (Niebylski et al. 1994) suggest bridge vector potential. Although highly anthropophilic (Gratz 2004), Ae. albopictus is an opportunistic species that will feed on birds and other vertebrates (Niebylski et al. 1994, Gubler et al. 2001, Gomes et al. 2003, Richards et al. 2006). Additional concerns involve the explosive population growth of Ae. albopictus since its discovery in the United States in Harris County, TX in 1985. Its tendency to breed in artificial containers expands the number of oviposition sites that are available to mosquitoes that inhabit urban and suburban areas where humans frequently supply these containers. Thus, surveillance programs are less effective against this species because they are unable to routinely treat private property where breeding sites may be abundant. Disease risk is a function of temporal and spatial patterns of vector habitats as well as the probability of interactions between humans and vectors (Dale et al. 1998). Urban infrastructure contributes to the maintenance of vector species by providing man-made habitats in place of natural ones. In highly urbanized settings, non-natural breeding sites may account for up to 93% of larval activity (Keating et al. 2003). Fortunately, habitats resulting from urban infrastructure may be more easily monitored through the use of Geographic Information Systems (GIS) as compared to traditional surveying techniques. Relating ecological, infrastructural, and entomological patterns is a valuable way to predict disease risk at different times of the year, and this strategy has been utilized to assign priority treatment in relation to dengue and malaria vector habitats (Thomson et al. 1997, Kazembe et al. 2006, Wen et al. 2006). Many counties in the United States have applied GIS tools to WNV surveillance in mosquito species. Studies varying by geographical region, however, have reached different conclusions regarding correlations among climate variables, land use, and disease risk (Pecoraro et al. 2007). In northwest Texas abundance of mosquito vectors on rural and agricultural lands was linked with precipitation, temperature and canopy cover (Bolling et al. 2005). In the Seattle, Washington area mosquito vector abundance was correlated with temperature, but not precipitation (Pecoraro et al. 2007). Conversely, in Rhode Island, WNV activity was associated with precipitation alone, and it is worth noting that the landscapes surveyed in this study were predominantly urban or suburban (Takeda et al. 2003). Multiple-county assessments in Colorado, Louisiana, Pennsylvania, and Nebraska demonstrated positive correlations among WNV infection, high temperatures, and agricultural land use (Miramontes et al. 2006). In Georgia, however, urbanization and not agriculture was related to WNV risk (Gibbs et al. 2006). In addition, seasonal mosquito collection data from municipal surveillance programs has been drawn upon to analyze ecological correlates of WNV risk in other regions of the U.S. (Allan et al. 2009). To date, no study has investigated relationships among mosquito abundance, WNV risk, and environmental factors in Virginia. Environmental inspectors in Henrico County, VA have documented trends in both avian deaths and WNV-positive mosquitoes that suggest increased arboviral activity in urban areas (Buchanan, personal communication). It is currently unclear, however, if the potential for WNV transmission by mosquito vectors in the eastern United States is related to landscape or climate factors or both. Our study seeks to determine whether certain climatic conditions or land uses are associated with increased arbovirus activity. To determine if the potential for WNV infection is related to climate factors, we analyzed abundance of vector species in relation to temperature and precipitation. To determine if the potential for WNV infection is related to urban landscape features, we examined three measures of development within our defined urban and suburban study regions: canopy cover, building footprint (Marzluff et al. 2001), and proximity to combined sewer overflow drainage infrastructure (Calhoun et al. 2007). We surveyed three potential WNV vectors, Aedes albopictus (Skuse), Culex pipiens pipiens (L.), and Culex restuans (Theobald), species that are consistently collected in the highest abundance and the only species to date to test positive for WNV in the jurisdiction. Our study shows that certain landscape variables identified using GIS are valuable for predicting the population growth of primary WNV vectors in Virginia and that temperature along with low precipitation are strong predictors of population growth. Henrico County, VA is located in the central region of the state to the north of the City of Richmond (latitude 37.342–37.726, longitude 77.158–77.671) and contains a mix of urban, suburban, and rural landscapes (Figure 1a). Maximum normal temperatures peak for Henrico County in mid-July averaging 30.5° C (NCDC 2009). Precipitation tends to be high during summer months, especially in July and August, with seasonal total averages approaching 50 cm from June through October. Active WNV surveillance in Henrico County began in 2002 with the formation of the Standing Water Initiative (SWI) within the municipal Public Works Environmental Division. Surveillance by SWI personnel includes active collection and testing of mosquito vector pools, testing of dead birds in response to citizen reports, and investigation of possible human cases. Our study drew upon the Public Works mosquito collection records from June through October 2005, 2006, and 2007. (a) Henrico County boundary showing the relative locations of Monument (urban) and Sandston (suburban) neighborhoods, (b) Monument area trap sites, and (c) Sandston area trap sites. Two study regions, one urban and one suburban as defined by GIS maps of existing land use, were selected by defining rectangles approximately 77 km2 in an area where visible clusters of avian deaths were reported during 2005–2007. These regions were classified according to major points along the gradient of urbanization as described by Marzluff et al. (2001). Urban sites were characterized by the presence of commercial and industrial land uses, high-density residential housing with interspersed public open spaces. Suburban sites were characterized by mostly low-density residential housing, higher percentage of public, open, or vacant lots with few interspersed commercial or industrial areas (Table 1). Sites were selected within the chosen regions if they met one or both of the following additional criteria: 1) captured mosquitoes tested positive for West Nile virus in at least one of the 2005, 2006, or 2007 trapping seasons or 2) the site was visited bi-weekly by environmental surveyors. A total of 37 sites were selected among urban sites (Monument neighborhood) (Figure 1b) and suburban sites (Sandston neighborhood) (Figure 1c). Although avian data were used to identify the regions as possible WNV hotspots, incidence of WNV in birds was not included in the following analyses because avian records existed only in response to citizen service requests and were not routinely collected. While mosquito trapping did occur in rural regions of Henrico County, VA, these sites were excluded from the study because they did not meet the selection criteria. Adult mosquitoes were collected from June through October using carbon dioxide light-baited traps and gravid traps consistent with CDC surveillance guidelines (CDC 2003). One of each trap type was placed at three sites nightly and allowed to run for 16–20 h to account for different temporal feeding patterns of Culex and Aedes females. Captured mosquitoes were immobilized upon return to the laboratory by placing them in a –4° C freezer or in the field using surplus dry ice from light-baited traps. Mosquitoes were then counted and identified to genus and species using the Key to the Mosquitoes of North Carolina and the Mid-Atlantic States (kindly provided by the North Carolina State University Agricultural Extension Service). Trap data for all vectors were pooled to calculate the mean vector abundance at each site (mosquitoes per trap per night) and to calculate the proportion of mosquitoes infected with WNV (Allan et al. 2009). Vector species were then pooled in groups of 10–50 individuals and taken to the Virginia Department of Health State Laboratory for WNV testing by PCR. Cx. restuans and Cx pipiens were further pooled as Culex spp. for West Nile virus testing as per the collection methods conducted by Kulasekera et al. (2001). Species-specific data were used only in analyses regarding landscape variables where flight range is species specific (see subsection GIS Observation and Mapping). All other statistical analyses use pooled vector abundance of Culex spp and Ae. albopictus as described above. To determine if mosquitoes use combined sewer overflow (CSO) stormwater inlet basins as larval habitats in the City of Richmond, larvae were sampled from ten inlet basins within city borders adjacent to the Monument Ave. neighborhood study site (Figure 1b). The City of Richmond and parts of Henrico County use CSO in which stormwater infrastructure receives combined storm runoff and wastewater effluent. In areas with CSO, inlet basins are present and are designed to hold water to prevent emanation of sewage-related objectionable gases into the surrounding neighborhood. Four independent larval samples were taken from each inlet basin using an aquatic larval surveillance dipper. After the basin manhole cover was removed, surveyors waited 5 min for disturbed larvae to return to the water's surface. One dip was taken from each of the four corners of the inlet basin. Samples were emptied into a bucket to prevent redundant counts and returned to the basin after counting. Larvae were sampled weekly from June 25, 2008 to September 29, 2008; presence of early instars (1st and 2nd), late instars (3rd and 4th), and pupae for all species were documented. Presence of egg rafts for Culex spp. was also documented. Presence of adults of either species flying or resting on inlet basin walls was noted. A fifth larval sample was periodically brought back to our lab and reared in an environmental chamber to confirm species identification. The Henrico County Public Works Department provided six-inch pixel resolution aerial photographs of Henrico County as well as Geographical Information Systems (GIS) shape files mapping environmental and infrastructural features of the study area. These shape files were constructed using ArcMap 9.2 (Environmental Systems Research Institute, Redlands, CA); they illustrated locations of land use, adult mosquito trap sites, building coverage, and inlet basins under the 1983 North American Geographic Coordinate System (Figure 1). Separate buffer zones corresponding to each species flight range, 200 m for Ae. albopictus (Bonnet and Worchester 1946) and 400 m for Culex. spp. (Vinogradova 2000), were created surrounding each trap site. Although recapture studies of female Ae. albopictus report a maximum dispersal of 400–600 m (Rosen et al. 1976, Niebylski and Craig 1994), Bonnet and Worchester (1946) provide evidence that the same species will disperse no more than 200 m where blood meals are available. After consultation with Randy Buchanan of Henrico County Public Works, we adopted 200 m as the local flight range for Ae. albopictus due to the high density of resting sites and potential hosts. Percent building cover (building footprint) was calculated and frequency of stormwater intake structures counted within each buffer zone. Temperature and precipitation data were obtained from the Richmond International Airport National Climate Data Center's (NCDC) weather station that is located within Henrico County, VA. Daily minima and maxima from June through October were used to generate weekly averages for 20 weeks of each year. In 2007 trapping did not begin until the second week of June, so in this year there were only 19 weeks included in our study. Total weekly precipitation was considered for these periods and values were adjusted by two weeks to account for the lag time between rainfall and adult mosquitoes trapped. Weekly temperatures and precipitation for the 2005, 2006, and 2007 mosquito collection seasons were compared to decade trends generated from 1995–2004 National Climate Data Center data. Canopy cover at each trap location was determined by establishing a 3 m2 grid where traps are routinely set. Forty-nine separate readings were taken within each grid using a spherical densitometer to calculate percent canopy cover. Since these readings were taken in 2008 after actual trapping dates, densitometer calculations were compared to 2003 aerial photographs to determine possible changes in vegetation due to construction or development. Proximity of traps to inlet basins was determined by consulting GIS shapefiles of infrastructure. The number of stormwater drop inlet structures within each buffer zone was counted. Building footprint was also determined by consulting GIS shapefiles. The total ground area covered by buildings was measured within each buffer zone to generate percent building coverage values. For all analyses, we defined abundance as the average mosquito catch per trap per night; vector abundance refers to the pooled abundance of both Culex spp. and Aedes albopictus. Chi square analysis was used to compare proportions of West Nile virus positive mosquito pools from Monument and Sandston neighborhoods (n = 37 sites). Logistic regression was used to examine correlations between pooled abundance and occurrence of West Nile virus positive pools at all trap sites (n = 37 sites). Logistic regression provided evidence that pooled vector abundance was strongly correlated with WNV positive pools (see Results). Thereafter, analyses were conducted using pooled vector abundance as the response variable. One-way ANOVA was used to compare vector abundance, temperature and precipitation among 2005, 2006, and 2007 trapping seasons. We also used one-way ANOVA to examine temperature and precipitation values among trapping and decade years. Comparisons between specific groups were further analyzed using Tukey Standardized Range test or linear contrasts. Linear regression was used to evaluate relationships between weekly abundance and weekly average temperature or weekly average precipitation. Multiple regression analysis was then performed to investigate temperature and precipitation as predictors of mosquito abundance over all years of data collection. We then applied a multivariate analysis to explore how temperature and precipitation related to mosquito abundance. Linear regression was used to examine the relationship between abundance and three landscape variables: proximity to inlet basins, percent building coverage, and percent canopy cover. Linear regression of landscape variables used species-specific abundance data; analyses were conducted separately for Ae. albopictus and Culex. spp. All variables were statistically analyzed using JMP 7.0 (SAS Institute Inc., Cary, NC) and illustrated using GraphPad Prism 4.0 (GraphPad Software Inc., San Diego, CA). Increased vector abundance was strongly correlated with positive WNV test results (n = 37, X2= 9.30, P = 0.0023). Mean pooled abundance was significantly different among 2005, 2006, and 2007 trapping seasons (F = 8.04, P = 0.0008) (Figure 2). Highest vector abundance was seen in 2005 and 2007 with a mean of 48 mosquitoes per trap per night each year. Mean vector abundance was significantly lower in 2006 with a mean of 27 mosquitoes per trap per night (Tukey P = 0.0031 and 0.0027 for 2005 and 2007, respectively) (Figure 2). Annual mean vector abundance (dark grey), temperature (black) and precipitation (striped) in 2005, 2006, and 2007. Mosquito larvae, pupae, and Culex egg rafts were observed in at least one of the sampled drop inlets for the entire duration of larval surveillance. First and second instars were generally found in the highest number of drop inlets (mean = nine inlets), followed by later instars (mean = seven inlets) and finally pupae (mean = four inlets) as shown in Table 2. Culex spp. adults were first observed in mid-July and followed a bimodal distribution for the remainder of the mosquito season, peaking in late July and again in mid-September. Ae. albopictus adults were first observed in late July and peaked in mid-September. Mean temperature for the twenty weeks from June through October was not significantly different among 2005, 2006, and 2007 (2005 mean = 25.3° C, 2006 mean = 24.1° C, 2007 mean = 24.7°, F=0.637, P= 0.53) (Figure 2) nor when analyzed along with decade data (F=1.51, P=0.119). However, compared to the decade average of 22.9° C, it was significantly warmer during trapping months in both 2005 (linear contrasts F=7.94, P=0.005) and 2007 (linear contrasts F=4.32, P=0.03), but temperatures were similar to decade trends in 2006 (linear contrasts F=1.99, P=0.158). While mean temperature did not exceed 32° C, daily maxima sometimes reached above 32° C. Temperature was positively correlated with vector abundance in 2005 only (r2 in 2005 = 0.26, P = 0.022). In 2005 vector abundance generally increased as temperatures increased in July (weeks 26–29) and began declining in October (weeks 39–41) as temperatures fell (Figure 3a). In 2006 vector abundance was not correlated to temperature (r2 in 2006 = 0.05, P = 0.34) but peaked early in the season and began declining in September (weeks 37–41) along with temperatures (Figures 3b). During 2007 vector abundance was also not correlated with temperature (r2 in 2007 = 0.0003, P = 0.94). In 2007 vector abundance peaked early in the season as it did in 2006, dropped sharply in late June (week 26), and remained relatively stable into October (week 40) (Figure 3c). a-c. Weekly temperature (black bars), precipitation (grey shaded area), and vector abundance (dark grey bars) from June through October in year (a) 2005, (b) 2006, (c) 2007. Week number corresponds to dates where week 20 = June 1–7, week 21 = June 8–15, etc. Mean precipitation from June through October was significantly different among 2005, 2006, and 2007 trapping seasons (2005 mean = 2.10, 2006 mean = 4.08, 2007 mean = 1.98, F=2.79, P=0.0425) (Figure 2) as it was when analyzed along with decade values (F=2.79, P=0.001). Compared to the decade average of 2.74 cm from June through October, there was suggestive evidence that precipitation for the trapping months was higher in 2006 (linear contrasts F=3.42, P=0.065), but precipitation values were similar to decade trends in both 2005 (linear contrasts F=0.729, P=0.393) and 2007 (linear contrasts F=1.04, P=0.308). There was also suggestive evidence that precipitation during June through October was greater in 2006 compared to 2005 (Tukey P = 0.061) and 2007 (Tukey P = 0.048) (Figure 2). Precipitation in 2005 was marked by peak rainfall early in the season (week 22) and again mid-season (weeks 29–32) followed by dry conditions (Figure 3a). Total rainfall for July 2005 (weeks 29–32) reached 23.6 cm, whereas late August through October 2005 (weeks 33–41) received a combined 6.7 cm. High vector abundance was also seen in 2005 (Figure 2) and was positively correlated with rainfall (r2 in 2005 = 0.24, P = 0.030). Precipitation data from 2007 also reflected dry conditions late in the season but a peak rainfall of 16.7 cm arrived after mid-season in September (weeks 34–36) (Figure 3c). Although annual values for vector abundance were lower in 2006 compared to 2007 (Figure 2), no significant correlation was found between weekly vector abundance and precipitation in either year (r2 in 2006 = 0.008, P = 0.693; r2 in 2007 = 0.07, P = 0.262) (Figure 3b-c). While results were not significant, analyses indicated that weekly mosquito abundance was negatively related to precipitation in both 2006 (Pearson correlation coefficient r =–0.09) and 2007 (Pearson correlation coefficient r =–0.27). Multiple regression analysis, along with multivariate technique, for both Ae. albopictus and Culex spp., provided evidence that temperature was a stronger predictor of mosquito abundance (r = 0.30; P = 0.02) compared to precipitation (r =–0.11; P = 0.40). Examination of species-specific data, however, revealed only modest evidence for partial correlation between temperature and abundance of Culex spp., while temperature correlations were not significant for Ae. albopictus (Table 3). While partial correlations between precipitation and abundance of both vector species were not significant, correlations were weakly negative (Table 3). Thirty-seven trap sites, of which twenty-seven were classified as "urban" and ten classified as "suburban," yielded 1,540 mosquito pools that were tested for WNV during 2005, 2006, and 2007 seasons. In 2005 we observed 22 positive pools, 12 positive pools in 2007, and six positive pools were observed in 2006. Chi Square Fisher's exact test supported that a significantly greater proportion of WNV positive pools originated from urban trap sites (3.9%) vs suburban trap sites (1.4%) (one-tailed P = 0.0094). In addition, urban sites were strongly associated with increased abundance of Culex spp. (t35= 3.20, P = 0.002). Urban or suburban classification was not associated with increased abundance of Ae. albopictus (t35=–1.37, P = 0.178). Summary statistics regarding these comparisons are shown in Table 4. Culex spp. abundance was positively correlated with the frequency of stormwater structures (r2= 0.21, P = 0.023), while there was suggestive but inconclusive evidence for an association between Culex spp. abundance and percent building coverage (r2= 0.09, P = 0.078). Culex spp. abundance was not correlated with canopy cover (r2= 0.0003 P = 0.942). Aedes albopictus abundance was not correlated with stormwater structure frequency (r2= 0.034, P = 0.281), building footprint (r2= 0.001, P = 0.821), or canopy cover (r2= 0.009, P = 0.569). Known WNV vectors are present in both urban and suburban landscapes within Henrico County, VA, and our study demonstrates that urban locations tend to produce a greater proportion of WNV-infected Culex species than suburban locations. This study is the first to document an association between urban sites and arbovirus activity in this region, a trend that is supported by increased abundance of primary vectors in conjunction with increased WNV positive mosquito pools. From our 2005 – 2007 mosquito collections from our urban and suburban study sites, we observed an average minimum infection rate (MIR) of 1.2 mosquitoes per 1,000 and 0.36 mosquitoes per 1,000, respectively. Although these values are below the MIR threshold for elevated concern in Virginia of five mosquitoes per 1,000 (VDH 2004), they show that infection rates can be highly variable even within a mosquito control district. Although our analysis demonstrated a correlation between arbovirus activity and urbanization, some landscape variables we chose to examine were unexpectedly poor predictors of vector abundance in Henrico County, VA. With the exception of stormwater structures in relation to Culex spp. abundance, neither building coverage nor canopy cover were strongly associated with Culex spp. abundance and none of the examined landscape variables were correlated with increased Aedes albopictus abundance. The strong correlation between stormwater structures and Culex spp. abundance does, however, provide insight into the primary vector dynamics that contribute to virus amplification when considered in context with precipitation trends. Our examination of the role of climate variables as contributing factors to vector abundance in Virginia supports modeled data in Florida that suggested WNV may proliferate under low rainfall conditions (Shaman et al. 2005). We observed increased WNV positive mosquito pools in 2005 and 2007, years that experienced high vector abundance and light rainfall. In 2005, we believe that warm temperatures sustained vector abundance over a long period of time within the mosquito season. Ample larval habitats, created early in the season by high precipitation, remained undisturbed over the remaining season due to lower than average rainfall. The opposite occurred in 2006 in which sustained precipitation over the course of the mosquito season may have flushed out existing habitats so that mosquito populations were reduced in that year. In 2007, mosquito numbers peaked early in the season but this was not explained by either temperature or precipitation data recorded for that year. Peak mosquito numbers in 2007, however, may have been sustained by the higher than average precipitation recorded in the previous 2006 year. Landesman et al. (2007) found that incidence of WNV was positively correlated with the previous year's precipitation in the eastern U.S., due possibly to larval habitats remaining from the previous growing season or due to large numbers of overwintering mosquitoes initiating population outbreaks the following year. Stormwater drop inlets provide tens of thousands of isolated Culex larval habitats that respond to precipitation events in this manner; by design these inlets hold water and may have contributed to Culex spp. abundance. For 2008 only, we sampled drop inlets in city areas adjacent to the Monument neighborhood study site and found that close to 40% of drop inlets produced pupae on any given sampling date. These drop inlets also produced high numbers of Culex adults in late July and again in September. It is likely that the drop inlets provided an ample source of undisturbed, stagnant water habitats for Culex during 2005 and 2007 mosquito seasons when warm temperature and low precipitation prevailed. The storm inlets examined in our study are part of the combined sewer-stormwater overflow system that is shared by Henrico County, the City of Richmond, and Chesterfield County. Our findings are consistent with previous literature documenting increased oviposition in other Culex spp. in the combined stormwater-sewer systems present in Atlanta, GA (Chaves et al. 2009). These sites are particularly attractive to Culex spp. due to the periodic influx of organic material. Combined sewer-stormwater systems are common in the southeast cities in the United States and should be a primary target for vector control. GIS maps of drainage infrastructure allow environmental surveyors to predict areas likely to experience high mosquito population growth. Drainage infrastructure in relation to building cover, zoning, and parcel maps may reveal where high production breeding sites intersect with human activity. In this manner, emerging surveillance programs can create a comprehensive picture of potential arbovirus activity and allocate limited resources to areas that produce mosquitoes likely to contact people. We expect Culex vector abundance and WNV activity to further increase with proximity to metropolitan areas, especially in cities with a high density of sewer-stormwater related inlets. Periodic mosquito surveillance conducted by the Virginia Department of Health Division of Zoonotic and Environmental Epidemiology in 2005 report a period MIR of 11.44 mosquitoes per 1,000 for weeks 27 to 41 in the City of Richmond (D. Gaines, pers comm). During this same period, Henrico County officials reported a county-wide seasonal MIR of 3.72 mosquitoes per 1,000. Characterizing disease risk on the neighborhood level is important especially in light of the high variability in MIR values within the county and in nearby urban centers. Such localized information will continue to facilitate the development of an effective control program in eastern U.S. cities geared toward controlling Culex vectors in the larval stages, but its use for Aedes albopictus control may prove more difficult. Our larval and adult sampling data suggest that both Culex spp. and Ae. albopictus proliferate in and near drainage infrastructure, but Ae. albopictus abundance was not correlated with any landscape variable tested in this study. Based on the anthropophilic nature of this species, the lack of correlation between urbanization and abundance is unexpected. It is possible that preference for small, man-made containers causes this species to be less predictable based upon larger-scale landscape features. Aedes albopictus proliferation may instead be based upon microclimates provided by private residents and future research should examine other variables related to Ae. albopictus interactions with human hosts (i.e., distance to human dwelling, presence of artificial containers, etc.). Its rapid expansion and high vector competence necessitate additional studies that explore a combination of surveying techniques to measure abundance and prevalence of larval habitats. Our data indicates that landscape variables available in GIS are valuable for predicting the population growth of abundant primary WNV vectors in Virginia. For Ae. albopictus, however, landscape characterization must be used in conjunction with remote-sensing and larval habitat surveys to recognize high risk locations. Mosquito collection and identification was conducted in collaboration with the Henrico County Department of Public Works Environmental Division, which provided CDC traps, gravid traps, adult identification, and larval sampling equipment. We thank Randy Buchanan and Lane Carr for information regarding Henrico County surveillance locations, techniques, and records. We also thank David Gaines (VA Department of Public Health) for sharing MIR data. University of Richmond undergraduate Jon Swanson and Michelle Fults (Richmond District Planning Commission) provided initial GIS consultation; University of Richmond undergraduate John Frey helped with canopy cover measurements. This paper is in partial fulfillment of the M.S. degree of J. D. Portions of this research were funded by University of Richmond Arts and Science Dean's funds to A.T.