By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Nearly 30% of emerging infectious disease events are caused by vector-borne pathogens with wildlife origins. Their transmission involves a complex interplay among pathogens, arthropod vectors, the environment and host species, and they pose a risk for public health, livestock and wildlife species. Examining habitat associations of vector species known to transmit infectious diseases, and quantifying spatio-temporal dynamics of mosquito vector communities is one aspect of the holistic One Health approach that is necessary to develop effective control measures. A survey was conducted from May to August, 2010 of the abundance and diversity of mosquito species occurring in the mixed-grass prairie habitat of the Smoky Hills of Kansas. This region is an important breeding ground for North America's grassland nesting birds and, as such, it could represent an important habitat for the enzootic amplification cycle of avian malaria and infectious encephalitides, as well as spill-over events to humans and livestock. A total of 11 species, belonging to the three genera Aedes, Anopheles, and Culex, was collected during this study. Aedes nigromaculis, Ae. sollicitans, Ae. taeniorhynchus, Culex salinarius, and Cx. tarsalis accounted for 98% of the collected species. Multiple linear regression models suggested that mosquito abundances in the grasslands of the central Great Plains were explained by meteorological and environmental variables. Temporal dynamics in mosquito abundances were well supported by models that included maximum and minimum temperature indices (adjusted R2= 0.73). Spatial dynamics of mosquito abundances were best explained by a model containing the following environmental variables (adjusted R2=0.37): ground curvature, topographic wetness index, distance to woodland, and distance to road. The mosquito species we detected are known vectors for infectious encephalitides, including West Nile virus. Understanding the microhabitat characteristics of these mosquito species in a grassland ecosystem will aid in the control and management of these disease vectors.
Monitoring vector communities is an integral part of disease surveillance and control programs. Nearly 30% of the emerging infectious disease events are caused by vector-borne pathogens with wildlife origins (Jones et al. 2008), whose transmission involves complex interactions among their vectors, the environment, and host species. For example, mosquito vector demography is closely associated with biophysical variables such as elevated summer temperatures that can reduce developmental time for pupal/larval life stages, increase mosquito abundances, and shorten extrinsic incubation periods of pathogens they carry. The timing and availability of amplifying host species can also change transmission dynamics of vector-borne pathogens.
Vector-borne diseases pose risks to the public health sector, livestock industry, and wildlife species of conservation concern. In the grasslands of the central Great Plains, multiple mosquito-borne infectious diseases have threatened public health, including Western Equine encephalitis (WEE), St. Louis encephalitis (SLE), and West Nile virus encephalitis. Infectious encephalitides and avian malaria exhibit enzootic avian cycles in grassland nesting birds, which have resulted in population declines (West Nile virus in Greater Sage-grouse, Centrocercus urophasianus, Naugle et al. 2004) and compromised fitness (avian malaria in Greater Sage-grouse, Boyce 1990) in these avian species of conservation concern. In addition to being affected by these infectious agents, avian epidemics may serve as a mechanism for the amplification of pathogens in the environment (Burkett-Cedena et al. 2001, Turell et al. 2001), leading to transmission events in humans and livestock. Hence, grassland habitats provide a unique interface for the transmission from the enzootic mosquito-avian cycles (involving grassland nesting birds) to mammalian spill-over hosts such as humans and livestock. Ultimately, examining the fine-scale habitat association of known vectors of infectious disease and quantifying spatio-temporal dynamics of mosquito vector communities in a habitat suspected of being a major center for pathogen spill over is an aspect of the holistic One Health approach and necessary to develop effective control measures.
Here, we focused on the distribution of mosquitoes in the grassland of the Smoky Hills, Cloud County, KS, during the period of nesting for grassland birds. The objectives were to 1) examine the spatio-temporal distribution of potential vectors of infectious encephalitides and avian malaria and 2) determine the underlying meteorological and environmental variables that give rise to peaks in vector abundance and diversity.
MATERIALS AND METHODS
The study site was located in north-central Kansas, in the Smoky Hills eco-region of Cloud County, KS (Figure 1). The Smoky Hills are tillable, with native prairies characterized by moderate to high fragmentation, row crop agriculture, and low intensity cattle stocking. The study area encompassed 283 km2 of fragmented prairie landscape; consisting of 73% grassland, 25% cropland, and a road density of 1.4 km/km2. Cultivated croplands included wheat, sorghum, soybeans, and corn. The climate was temperate, with moderate rainfall and annual precipitation during the sample year of 883 mm, of which nearly half could be attributed to precipitation from May through July, 2010.
In 2012, the Kansas Department of Health and Environment (KDHE) reported the highest number of human WNV infections since 2003. There were three outbreaks of mosquito-borne infectious disease (WNV, KDHE 2012) within ten counties surrounding the study site. Human cases of infectious encephalitides occurred in 2003, 2008, and most recently in 2012. The last human incidence report in Cloud County occurred in 2003 (West Nile virus encephalitis, KDHE 2003).
Using geographic information system software, we randomly selected twenty collection sites within grassland habitat (Figure 1). Adult mosquitoes were collected in Center for Disease Control miniature light traps (J.W. Hock, Gainesville, FL) weekly from May to August, 2010. Light traps were baited with 1 kg dry ice, powered by 6-V lantern batteries, and hung 1.5 m above the ground. The traps were deployed before dusk and retrieved the following morning. Collected mosquitoes were transported to the laboratory in ice chests filled with dry ice and subsequently stored at −20° C until identification. Adults were identified to species level using dichotomous keys (Pratt and Stojanovich 1961). Field-collected specimens of Culex species were often difficult to differentiate due to missing morphological characters. Differentiation between Culex restuans and Culex pipiens was achieved by the presence of two small white-scaled round spots on the scutum (80% success rate, Apperson et al. 2002).
Meteorological and environmental data
Daily weather data, including precipitation, wind speed, and minimum and maximum temperatures were obtained from the National Oceanic and Atmospheric Administration (NOAA) at Concordia Blosser Municipal Airport (N 39°32.949’ W 097.39.133’), 12 km north of the study site. We calculated mean weekly precipitation, temperature (minimum, maximum), and wind speed and accounted for lagged responses to precipitation and temperature by including conditions one and two weeks prior to May 17th (two weeks before sample season) to August 2nd, 2010.
Environmental data were acquired from Kansas Geospatial Community Commons (http://www.kansasgis.org) and processed in Arc Info 10 (Environmental Systems Research Institute, Radlands, CA) for geospatial analysis and data extraction. Landcover analysis was conducted using the 30 m resolution of the 2005 landcover map of Kansas (Whistler et al. 2006) depicting the following landcover classes: urban industrial/commercial, urban residential, urban open land, urban woodland, urban water, cropland, grassland, CRP (Conservation Reserve Program), woodland, and water. Furthermore, we used the 1999 National Elevation Dataset (U.S. Geological Survey, EROS Data Center) and roadway dataset that combined the 2006 Kansas State and Non-State Road System dataset (Kansas Department of Transportation: Bureau of Transportation Planning). Each landcover data set was aggregated to 30 m grain size prior to geospatial analysis. Using this dataset, we estimated five variables at each mosquito trapping location: distance to agriculture, distance to woodland, distance to water, distance to road, and distance to edge. Characteristics used to evaluate the area surrounding the trapping locations were: curvature and topographic wetness index (TWI), which were analyzed using the 1999 National Elevation Dataset and Spatial Analyst tools. Curvature describes local variations in surface topography, for example a positive curvature indicates a convex surface, while a negative curvature indicates a concave surface likely to result in stagnant water bodies. TWI describes the predicted soil moisture pattern and is calculated as the natural logarithm of the ratio between local upslope contributing area and slope (Pathak 2010). These landscape characteristics were evaluated at the attractiveness range of CO2-baited mosquito traps by applying a 30 m radius buffer to each trapping location. The attractiveness range of CDC CO2-baited light traps is somewhat ambiguous, however it has been shown that CO2 attracts mosquitoes from distances of <30 m and is species dependent (Service 1993). Thus, a 30 m buffer was selected to encompass mosquito habitat around trapping locations.
Statistical analyses were performed on abundance and diversity data (Shannon Index, H’) of the mosquito community to determine which meteorological and environmental variables were most influential on these data. Several steps were taken to avoid overparameterizing of biophysical models of mosquito community data. Meteorological and environmental variables were examined for colinearity using Pearson's correlation matrices and redundant or highly correlated variables (r>0.90) were excluded from further analyses. Prior to model construction, variables were tested for normality using Kolmogorov-Smirnov tests. Relative abundance and TWI showed significant departures from normality and were arcsin and log10 transformed, respectively.
To investigate the relationship of mosquito abundance or diversity with meteorological and environmental variables, multiple linear regression analysis was performed to fit the dependent variable, either mosquito abundance or diversity, to the independent meteorological and environmental variables. The effects of meteorological variables on mosquito abundance or diversity were examined using weekly averages of precipitation, wind speed, maximum temperature, and minimum temperature. To account for lagged responses to precipitation and temperature, conditions one and two weeks prior were included as independent variables. Mosquitoes were sampled weekly for a total of nine weeks. Each week, estimates of mosquito abundance and diversity were averaged over the 20 sampling locations. Due to the small sample size, data were not split for cross-validation. The full models included six weather variables and an intercept for a total of seven parameters. Candidate models for mosquito abundance and diversity included all possible combination of weather variables (26 = 64 models for each dependent variable). All six weather variables had been previously shown to be linked to mosquito abundance or diversity; hence each possible combination and interaction terms of these variables was included.
We examined separately the effects of environmental variables on mosquito abundance and diversity; variables examined included curvature, TWI, distance to agriculture, distance to water, distance to woodland, distance to road, and distance to edge. Twenty locations were sampled, and abundance and diversity of mosquitoes was averaged over time (nine sample occasions). Data were not split for cross-validation due to the small sample size. The full models included seven environmental variables, and an intercept for a total of eight parameters. Candidate models included all possible combinations of environmental variables (27 = 128 models for each dependent variable). We included combinations of binary interaction terms for all seven environmental variables.
The most parsimonious models were selected using Akaike's Information Criterion with small sample size bias adjustment (AICc) (Burnham and Anderson 1998),
where k is the number of model parameters and N is the number of sample points, AICc has been shown to be superior to AIC. All models within 2 units of the minimum AICc value have substantial support, and should be considered for inferences (Burnham and Anderson 1998).
To determine biologically significant variables we calculated the cumulative AICc weights (Flanders-Wanner et al. 2004) of 2n models created for the analysis of mosquito abundance vs meteorological/environmental variables. The cumulative weight of a variable is calculated by summing up the AICc model weights of model (2n) containing that variable:
Here, wi are Akaike weights for model i, Δi is the difference between best fitting model, and model i. The denominator is the sum of the relative likelihoods for all candidate models.
Mosquito community description
The data collected in this study provide a description of the grassland mosquito community in the Smoky Hills eco-system in Cloud County, KS, during the nesting season of the grassland bird community. Mosquito samples were collected for nine weeks (once per week) at 20 sample sites from May through July (Figure 1). A total of 12,861 individual mosquitoes were collected, of which 11,223 (87.3%) could be identified to species level. The remaining 12.7% could not be identified due to missing or damaged morphological characteristics of field-collected samples. The mosquitoes captured in this study belonged to three genera: Aedes, Culex, and Anopheles, representing 11 taxa. Aedes was the most abundant genus, comprising 87.6% of the total collection, followed by Culex (12.1%) and Anopheles (0.3%). Of all taxa, Aedes sollicitans was the most abundant species collected during this study at 49.7%. Aedes nigromaculis (21.5%) was also very abundant throughout the study. Other common species included Aedes vexans (9.0%), Culex tarsalis (7.4%), Aedes taeniorhynchus (6.6%), and Culex salinarius (3.6%). Species encountered at lower abundances were Culex pipiens, Culex restuans, Aedes dorsalis, Aedes stimulans, and Anopheles species. Of the previous 35 mosquito taxa (DeMoss Hill 1939, Edman and Downe 1964, Janovy 1966, Lungstrom 1954, Lungstrom and Sooter 1961) reported in Kansas, nine were collected in this study. The collections include two previously unreported species in Kansas: Aedes stimulans and Aedes taeniorhynchus.
Temporal analysis revealed similar population fluctuation in all three genera. Standardized mosquito abundances indicated early season peaks in May and June followed by a decline in the last quarter of June and resurgence during July (Figure 2).
Aedes species exhibited small peaks in abundance mid-June and reached their highest numbers in July, with the exception of Ae. vexans. Ae. vexans reached peak abundances in mid-June and exhibited a minor peak in July. Ae. sollicitans was the most abundant species from June through the end of July and exhibited a small peak in abundance in the middle of June and achieved a major peak in the middle of July. Ae. nigromaculis, the second most abundant species, was similar to Ae. sollicitans; this species reached a small peak in the middle of June and a maximum in July. Ae. taeniorhynchus abundances were low throughout May, and June. In July, population numbers increased, reached maximum abundances in the third quarter of July, and remained stable until the end of the sample season. Culex species were encountered at lower abundances than Aedes species; however, Cx. tarsalis was the most abundant species collected in May. Abundances of Cx. tarsalis decreased in June and collection numbers remained small throughout the sample period. Cx. salinarius abundances remained low until the end of June and increased until July. Species within the genus Anopheles were the least abundant during this study; numbers remained low throughout the season and reached peak abundances in mid-July and declined until the end of the sampling season.
Throughout the sampling season we trapped a total of 29 anopheles species, hence this genus was encountered at the lowest abundances.
Associations between meteorological/environmental variables and mosquito abundance/ diversity
Multiple linear regression models testing the association among meteorological variables and mosquito abundance or diversity were reduced from ten to six weather variables, due to the high correlation between sample week temperature indices, and lagged temperature indices (a priori significant threshold, r>0.80). Analysis of the relationships among meteorological variables and mosquito abundance produced two plausible models based on their AICc values (Table 1). None of the most plausible models contained precipitation or wind variables. Temperature variables explained the temporal variation in observed mosquito abundances. The two best-fit models had an adjusted R2 of 0.73 and included one temperature variable. Minimum temperature (t=4.71, p=0.0001), and maximum temperature (t=4.47, p=0.0001) were positively correlated with mosquito abundance. Cumulative AICc weights indicated that minimum and maximum temperature were biologically significant variables, accounting for 69% and 30% of the AICc weights, respectively (Table 2). Unlike the abundance data, no best-supported model was found for the mosquito diversity index data. None of the meteorological variables used in this study explained the spatial variation in observed mosquito diversity.
Table 1. Candidate models used to fit the dependent variable, mosquito abundance, to independent meteorological variables.
Variables in the model
No. of parameter
aAICc = Akaike's Information Criterion with small-sample bias adjustment (Burnham and Anderson 1998). bAICc weight = percent of total weight from 128 models that can be attributed to the specified model. cTmin = minimum temperature during the sample week. dTmax = maximum temperature during the sample week. ePmean = precipitation during the sample week. fPmean(2) = precipitation two weeks prior to sample week, to account for lagged responses.
Tmin, Pmean, Pmean(2)f
Table 2. Cumulative AICc weights analysis for meteorological and environmental variables hypothesized to influence mosquito abundance in the Smoky Hills of Cloud County, KS, 2010.
Cumulative AICca weightb
Cumulative AICc weight
aAICc = Akaike's Information Criterion with small-sample bias adjustment (Burnham and Anderson 1998). bCumulative AICc weight of a variable = the percent weight attributed to models containing that particular variable. Cumulative AICc weight is calculated as the sum of AICc model weights containing that variable. cTWI = Topographic Wetness Index, calculated as the natural logarithm of the ratio between local upslope contributing area and slope, and describes the predicted soil moisture pattern. dcurvature = is a measurement of rate-change of the slope per unit distance and may be an indicator for of aquatic habitat stability.
distance to road
distance to woodland
precipitation two weeks prior
distance to edge habitat
precipitation one week prior
distance to agricultural field
distance to water source
Multiple linear regression models testing the associations among environmental variables and mosquito abundance and diversity were constructed with the full set of environmental variables (n=7). The Pearson's correlation matrix did not indicate significant correlations among these variables (a priori significant threshold, r>0.80). All plausible models contained TWI but excluded the following distance variables: distance to agriculture, distance to water, and distance to edge habitat (Table 3.). The top model had an adjusted R2 of 0.37 and included the following variables: TWI, distance to road, distance to woodland, and curvature. TWI and distance to road were significantly correlated with mosquito abundance; TWI had a positive correlation (t=3.51, p=0.0001), while distance to road had a negative correlation (t=2.28, p=0.04). Distance to woodland and curvature showed a weak correlation with mosquito abundance. The cumulative AICc weights analysis indicated that TWI, distance to road, distance to woodland, and curvature were biologically significant variables, accounting for 81, 52, 45, and 41% of the AICc weights, respectively (Table 2.). Unlike the abundance data, we found no best-supported model for the associations among environmental variables and mosquito diversity. None of the environmental variables used in this study explained the spatial variation in observed mosquito diversity.
Table 3. Candidate models used to fit the dependent variable, mosquito abundance, to independent environmental variables.
Variables in model
No. of parameters
aAICc = Akaike's Information Criterion with small-sample bias adjustment (Burnham and Anderson 1998). bAICc weight = percent of total weight from 128 models that can be attributed to the specified model. cCurvature = is a measurement of rate-change of the slope per unit distance and may be an indicator for of aquatic habitat stability. dTWI = Topographic Wetness Index, calculated as the natural logarithm of the ratio between local upslope contributing area and slope, and describes the predicted soil moisture pattern (ESRI, 2010). eDistwl = distance to the closest woodland, calculated using Euclidean distance (Arc Info 10). fDistr = distance to the closest road, calculated using Euclidean distance (Arc Info 10).
Curvc, TWId, Distwle, Distrf
TWI, Distwl, Distr
Curv, TWI, Distr
The variation in abundance of host-seeking mosquitoes in Cloud Country, KS, showed a clear temporal pattern that was explained by meteorological data. Multiple linear regression suggested a positive association between mosquito abundance and temperature indices (minimum/maximum temperature). Minimum and maximum temperature indices were highly correlated (Pearson correlation coefficient r=0.80), making it difficult to discern their relative impact on the regression models of mosquito abundance. However, the cumulative AICc weights analysis suggest that maximum temperature (Cumulative AICc weight = 0.69) contributed more to the models than minimum temperature (Cumulative AICc weight = 0.30). Increased environmental temperatures likely drove mosquito abundance by increasing metabolic rates, reproductive output, and host-seeking behavior of these vectors (Shone et al. 2006), which has previously been shown in laboratory and field studies (Chuang et al. 2011).
In addition to the seasonal pattern observed in this study, the data indicate spatial variation of mosquito abundances across sample sites. Among the environmental variables considered, only TWI showed a strong positive association with mosquito abundances across sites. TWI is a wetness index that estimates a surface's potential to accumulate water based on the ratio between upslope contributing area and slope. Hence, the positive correlation with mosquito abundance may be due to the increased availability of larval habitats in areas with a higher TWI. This index has been widely used in hydrological studies and has previously been shown to predict spatial variation in mosquito communities (Clennon et al. 2010, Cohen et al. 2008, Shaman et al. 2006); hence it is an important variable to include in future habitat models. Furthermore, we observed a weak negative association with distance to road. The distance to roads has been observed previously to be an important variable to determine mosquito distribution (Khatchikian et al. 2011); hence habitat degradation and increase in road density may have implications on the spread of vector-borne diseases.
The transmission Of infectious encephalitides (WEE, SLE, WNVE) and avian malaria is maintained in an enzootic transmission cycle between birds and mosquitoes allowing for the amplification of the pathogens in the environment. In the mixed-grass prairie of the Smoky Hills this amplification cycle is likely to occur during the nesting season (early summer) of grassland nesting birds. During this time, females spend extended periods of time in a brooding position which reduces defense behavior towards mosquito attacks and increases disease transmission, while nestlings lack behavioral and morphological defenses to ward off mosquito parasitism, resulting in peak parasitemia during the nesting season (Burkett-Cadena et al. 2011, Valkiunas 2005). Cx. tarsalis, an important vector of encephalitides including WNV, SLE (Sirigireddy et al. 2006, Hammon and Reeves 1943) and avian malaria (Janovy 1966) in the Midwest, was the dominant mosquito species during this period. A peak in abundance occurred in the last quarter of May, during which this mosquito exhibited an ornithophilic feeding preference (Kent et al. 1943). As the season progressed, a shift in the mosquito community composition towards opportunistically and mammalian feeding Aedes species occurred. This shift in host-feeding pattern has been implicated in increased intensity of human and livestock epidemics of encephalitides, particularly following high early season abundances of ornithophilic feeders. Likely bridge vectors were Ae. sollicitans and Ae. vexans, but although both species are susceptible and capable of transmission to livestock and humans, their role in the enzootic cycle is low (Molaei and Andreadis 2006). During the sample year, no human cases of infectious encephalitides were reported in the study region, which may be due to the observed low amplitude peak of early-season ornithophilic mosquito species. Low abundances of these species have been shown to result in decreased tangential transmission to humans (Weaver and Reisen 2010).
Vector-facilitated spread of infectious diseases is of particular concern in the Great Plains, because this region represents the largest North American grassland habitat and is the breeding ground for grassland nesting birds, which as a group are experiencing sharp population declines (Rahmig et al. 2009). While habitat loss and degradation are implicated in the imperilment of these species, increased pathogen load may exacerbate the problem and quicken the decline. In grassland nesting birds, infectious encephalitides and avian malaria exhibit an enzootic avian cycle, which has been shown to result in large scale population declines (West Nile virus in Greater Sage-grouse, Centrocercus urophasianus, Naugle et al. 2004) and compromised fitness (avian malaria in Greater Sage-grouse, Boyce 1990). Disease transmission in the central Great Plains is not only of concern due to the conservation status of grassland nesting birds but also the association of important disease vectors (Culex spp.) with rural landscapes, specifically those dominated by perennial grasses used for grazing and hay production (Chuang et al. 2011), such as the mixed-grass prairie of the Smoky Hills. In Kansas, infectious encephalitides and avian malaria disease agents have been detected in field-collected mosquito samples (WNV, Sirigireddy et al. 2006; Plasmodium, Janovy 1966). To implicate specific mosquito species in the spread of infectious diseases, environmental conditions conducive to increased vector abundance, vector competence and host preference need to be evaluated.
There is an increasing recognition that the dynamics of infectious disease transmission is a result of the complex interplay between human animal and environmental health (Leung et al. 2012, Palatnik-de-Sousa and Day 1961). The One Health approach may help decipher these interactions among pathogens, arthropod vectors, environmental influence, reservoirs of infectious (e.g., avian species), and other susceptible hosts (e.g., livestock, humans) of infectious encephalitides and avian malaria in the Great Plains.
Examining the temporal and spatial dynamics of mosquito communities, when disease transmission is likely to occur, is integral for implementing surveillance programs and control measures. The purpose of this study was to establish general seasonal distribution and population patterns across a grassland ecosystem of various mosquito species that might be implicated in the spread of infectious disease and relate it to local transmission dynamics. While the information presented here contributes to the existing survey of mosquito communities in the Great Plains, it is important to consider that seasonal fluctuations of mosquito communities are variable across years, affecting both the abundance of individual species and the community composition, and their subsequent response to meteorological and environmental variables.
The authors gratefully acknowledge Dr. Ludek Zurek for providing trapping equipment and his expertise in the sampling of mosquito communities.