An accurate assessment of the local mosquito community is necessary to address the risk posed by mosquito-associated pathogens and guide abatement programs (Fouet & Kamdem 2019; Rund et al. 2019a). Entomological monitoring can detect vector and nuisance species of interest that contribute to accurate abundance estimates in order to identify changes in community composition. This data may provide valuable information toward effective mosquito control intervention strategies and timely public health awareness campaigns. Robust and sustainable mosquito surveillance strategies address the needs of the local community, are cost-effective, and maximize resources. An important step in this process is the evaluation of current monitoring strategies to ensure they are capable of addressing new challenges (e.g., range expansions, exotic invasive species) and detecting the full diversity of resident mosquito species while optimizing sampling effort.

In the state of Florida, mosquito control is primarily conducted by independent districts, as well as city and county funded programs. Therefore, surveillance methodology varies greatly across the state. One challenge to comparing surveillance data from mosquito control districts includes differences in trapping approaches and attractants used between individual programs. Quantifying the differences in mosquito abundance and community composition as it relates to surveillance methodology provides a first step toward understanding and communicating cross-district arbovirus risk over continuous geographic areas.

Generally, carbon dioxide and octenol are attractants used to increase CDC light trap collections or to target specific species (Newhouse et al. 1966; Takken & Kline 1989). Therefore, we analyzed collection data from 2 adjacent Florida mosquito control districts that used light traps baited with either of these 2 attractants in order to compare mosquito abundance and species composition. In addition, we wanted to visualize seasonal distributions for individual species, evaluate sampling efforts for each attractant, and quantify the potential for attractant bias across traps between districts. We hypothesize that (1) mosquito community composition will differ between districts from traps baited with each attractant, (2) a greater sampling effort will be required to capture the diversity of mosquito species in an area using traps baited with either attractant, and (3) traps baited with octenol will bias strongly toward specific species, reducing the diversity of species and associated trapping effort. Results from these comparisons will provide novel opportunities to describe mosquito distributions over continuous geographic areas for adjacent mosquito control districts when multiple attractants are used.

Materials and Methods

We obtained light trap mosquito surveillance data at 6 locations from Anastasia Mosquito Control District in St. Johns County, Florida, USA (30.04743°N, 81.54668°W; 30.12702°N, 81.62142°W; 30.15853°N, 81.36582°W; 30.15321°N, 81.39275°W; 30.02335°N, 81.37789°W; and 29.97751°N, 81.48103°W) and 4 locations from the city of Jacksonville Mosquito Control in Duval County, Florida, USA (30.5437200°N, 81.7284300°W; 30.1999900°N, 82.0083100°W; 30.4020500°N, 81.6430300°W; and 30.2121800°N, 81.6245100°W) (Fig. 1). Adult mosquitoes were sampled weekly from wk 19 to 45 (approximately mid-May to mid-Nov) in 2017 and 2018 using either CDC suction light traps (John W. Hock Co., Gainesville, Florida, USA) or American Biophysics Corporation light traps with light on (American Biophysics Corporation, East Greenwich, Rhode Island, USA). Traps set in St. Johns County were baited with octenol (BioSensory Inc., Putnam, Connecticut, USA) and those set in Duval County were baited with 2 kg of dry ice in 0.5 gal Igloo coolers (John W. Hock Co., Gainesville, Florida, USA). All traps were set in permanent locations and suspended about 1 m from the ground using a shepherd's hook. Traps were set between 3:00 P. M. and 5:00 P. M. and collected after 18 to 20 h. Mosquito collections were transported from the field back to their respective offices and identified to species by mosquito control personnel using the keys of Darsie and Morris (2003), Darsie and Ward (2005), and Burkett-Cadena (2013).

Fig. 1.

Map of Florida and location of trap sites.

Results

A total of 32,595 female mosquitoes comprising 9 genera and 30 species were included in the final data set (Table 1). These data are available as supplementary material for download in MIReAD format (Rund et al. 2019b) (Supplementary Table 1). We identified 12 potential or actual vectors known to be involved in the maintenance and spillover cycles of current and re-emerging zoonoses using the publications of Wellings et al. (1972), Bigler et al. (1972), Blackmore et al. (2003), Ortiz et al. (2005), and Foster and Walker (2019) (Fig. 2). The 4 most abundant mosquito species in traps baited with dry ice were Aedes infirmatus Dyar & Knab, Culex erraticus (Dyar & Knab), Culex nigripalpus Theobald, and Psorophora columbiae (Dyar & Knab) (all Diptera: Culicidae); those from octenol were Aedes atlanticus Dyar & Knab, Ae. infirmatus, Anopheles crucians Wiedemann, and Cx. erraticus (all Diptera: Culicidae) (Table 1). Culex species together made up 61% and 17% of our total collections for traps baited with dry ice and octenol, respectively. Culex nigripalpus was the most abundant Culex species in the dry ice baited traps, and Cx. erraticus similarly was the most abundant in octenol baited traps. Culex coronator Dyar & Knab and Culex salinarius Coquillett (both Diptera: Culicidae) were not collected in octenol baited traps.

Aedes atlanticus and An. crucians showed a clear affinity for octenol. Aedes atlanticus, which was not collected in dry ice baited traps, made up 38% of the octenol-baited collections and An. crucians made up 27% of the octenol-baited collections. Aedes aegypti (L.) and Aedes albopictus (Skuse) (both Diptera: Culicidae) were not well-represented in the data set, making up 4% of collections from traps baited with dry ice and < 1% of collections with octenol. Other mosquito species represented approximately 30% of the remaining collections from dry-ice and 15% of octenol baited traps (Fig. 2).

Several potential or actual vector species were collected in large numbers (> 1,000 per trap night) throughout the sampling period that included Ae. atlanticus, An. crucians, Cx. erraticus, and Cx. nigripalpus (Fig. 3). Aedes atlanticus, An. crucians, and Cx. erraticus abundance was greatest from mid-Jun to early Jul from octenol collections (Fig. 3). Culex nigripalpus and Cx. erraticus abundance was greatest from late-Jul to mid-Aug from dry ice collections (Fig. 3).

We found that dry ice and octenol collections shared 20 of the 30 mosquito species captured in traps supporting our hypothesis that the mosquito community would differ between traps with different attractants. This conclusion was based on the results of a low Morista-Horn similarity index (0.24). The Shannon diversity estimates indicated that mosquito collections in traps baited with dry ice and octenol contained about 8 and 6 common species, respectively, and Simpson diversity estimates suggested the presence of about 4 dominant species from each attractant (Table 2). Rarefaction/extrapolation curves indicated that sufficient trapping effort exists for traps baited with octenol, but additional trapping effort is needed to capture the full diversity available for traps baited with carbon dioxide (Fig. 4).

Table 1.

Overall mean mosquito abundance per trap night from dry ice and octenol CDC baited suction light traps from 2 mosquito control programs in northeastern Florida from Julian date 19 to 45 during 2017 and 2018.

Fig. 2.

Stacked bar plots of vector species of importance.

Fig. 3.

Phenology of mosquito taxa captured in light traps baited with dry ice (grey) and octenol (black) from northeastern Florida.

Results of the partial redundancy analysis revealed that a large proportion of the variance was explained by the conditional latitude and longitude variables (48.2%) of trap location. Results from the permutational ANOVA on the redundancy analysis confirmed a significant linear trend in the data (F = 3.260; df = 2; P = 0.006). Also, global dbMEM analysis found no evidence of remaining spatial autocorrelation in the data (P = 0.995). These results indicated that the inclusion of latitude and longitude of trap locations alone as conditional variables in the analysis was sufficient to account for potential bias resulting from the linear trend in the data. Results of the partial redundancy analysis indicated that the constrained variable attractant explained 14.1% of the total variance in the data set (R² _adj = 0.101; F = 2.253; df = 3; P = 0.047), and the unconstrained variance was equal to 37.6%. The first redundancy analysis axis (RDA1) explained 27.3% of the constrained variance (F = 2.253; df = 1; P = 0.035), and the first principal component axis (PC1) explained 30.5% of the unconstrained variance (Fig. 5).

Table 2.

Summary of diversity measures (± SE) and sampling effort from light traps baited with dry ice or octenol in northeastern Florida from Julian date 19 to 45 in 2017 and 2018.

Fig. 4.

Individual-based rarefaction curves for light traps baited with dry ice and octenol at 10 locations in Duval and St. Johns counties, Florida, USA during 2017 and 2018. The 95% confidence intervals are shown as the shaded regions.

Fig. 5.

Redundancy analysis (RDA) tri-plot of model output. Individual sites are represented as a shape (square or circle) to denote the attractant (dry ice or octenol) used at that location. The triangles represent the centroids of the attractants (i.e., the explanatory variables).

The coordinates in the ordination space represent the weighted sum of the species scores for each location. Each species name represents the tip of a vector extended from the origin, and the coordinates are derived from fitted values of each corresponding species. The location of species in the ordinational space relative to the attractant centroids indicates the strength of the correlation. The cosine of the angle between any 2 vectors reflects their linear correlation (Borcard et al. 2011). Correlations between vectors and centroids are approximated by right-angle projections of the centroid onto the response variable vector, where the final projection approximates the correlation between the 2 variables (Borcard et al. 2011).

The partial redundancy analysis tri-plot of the model output is shown in Figure 5. The large cluster of taxa in the center of the plot revealed that most taxa are heterogeneously dispersed among the 2 attractants. We observed Ae. atlanticus and Cx. erraticus to be associated strongly with octenol and Ae. infirmatus and Cx. nigripalpus with dry ice. We did not observe sample site locations to cluster by attractant.

Discussion

We observed substantial differences in the diversity and proportion of mosquito species related to attractant. Traps baited with octenol attracted Ae. atlanticus, An. crucians, and Cx. erraticus in larger numbers, and captured a greater proportion of species compared with dry ice traps, but dry ice traps yielded the greatest diversity of species and were more efficient at attracting Culex mosquitoes than octenol traps, except for Cx. erraticus.

As mentioned earlier, species richness differed between collections baited with the 2 attractants. Traps baited with octenol approached a clear and defined asymptote, indicating that the sampling effort using this attractant was sufficient to capture the full diversity of species that were going to be captured using this approach. The dry ice rarefaction curve did not approach an asymptote, indicating a need for increased sampling effort in order to capture the full diversity of species available using this approach (Fig. 4). Despite the need for additional sampling effort when using dry ice, a greater number of species were captured. The large number of single species (referred to as singletons) in mosquito collections from traps baited with dry ice may have contributed to larger confidence intervals in the resultant rarefaction curve than one produced for octenol.

Collectively, the results of our study support the need for a diverse sampling approach to describe the full community of mosquito species in northeastern Florida, and to obtain accurate estimates of abundance for actual and potential vector species of importance. Although many mosquito control programs employ multiple surveillance approaches, differences in trapping methodology used in adjacent counties can introduce challenges when determining overall arbovirus risk across contiguous geographic areas. Quantifying and visualizing differences between mosquito community composition and abundance, required sampling effort, and the potential for bias in trapping methods using different attractants will lead to improved understanding of arbovirus risk when relying on surveillance data from multiple sources.