https://orcid.org/0000-0001-6069-1198
Figures
Abstract
How landscape composition and configuration impact the distribution of multi-vector and multi-host mosquito vector-borne disease systems, such as West Nile virus (WNV), remains challenging because of complex habitat and resource requirements by hosts and vectors that affect transmission opportunities. We examined correlations between landscape composition and configuration and 2018 WNV sentinel chicken seroconversion in Florida, USA across the state and within five National Oceanic Atmospheric Administration (NOAA) bioclimatic regions to understand strength and variation of landscape effects during an elevated transmission year. Although few landscape studies have examined WNV in Florida, we expected higher percentages of residential or medium-developed landscapes and more fragmented landscapes would be positively correlated with WNV seroconversion owing to the main mosquito vector habitats and avian host distributions. However, we expected to find variation in the importance of forest, wetland, and agriculture landscapes across bioclimatic regions in the state. WNV seroconversion rates were calculated using Florida 2018 Department of Health WNV sentinel chicken seroconversion data from 187 flocks maintained by mosquito control programs. Percent land cover and edge density metrics were calculated for multiple land cover classes and within multiple buffer distances from chicken coops using 2019 National Land Cover Data. We used binomial generalized linear mixed effects models to calculate the importance of landscape metrics to WNV seroconversion. We found no statewide predictors of seroconversion, but as expected, the importance of landscape varied across regions. In the north-central part of the state, we found higher seroconversion in less populated suburban areas while higher seroconversion in south-central Florida was correlated with fragmented forested areas within 0.5 km of coops and intact woody wetland areas within 2 km of coops. This work corroborates previous findings that consistent landscape predictors of WNV are difficult to identify across broader geographic areas and sets the stage for additional work that incorporates climate and landscapes interactions for a greater understanding of WNV ecology in this geographic region.
Citation: Tavares Y, Day J, Giordano BV, Eastmond B, Burkett-Cadena N, Guralnick RP, et al. (2024) Regional variation in the landscape ecology of West Nile virus sentinel chicken seroconversion in Florida. PLoS ONE 19(10): e0305510. https://doi.org/10.1371/journal.pone.0305510
Editor: Sana Eybpoosh, Pasteur Institute of Iran, ISLAMIC REPUBLIC OF IRAN
Received: May 31, 2024; Accepted: September 20, 2024; Published: October 25, 2024
Copyright: © 2024 Tavares et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Georeferenced sentinel chicken seroconversion data is available upon request through the Florida Department of Health Arbovirus Surveillance program upon agreement from participating Florida mosquito control programs through a memorandum of understanding. The authors did not receive special privileges in accessing the data that other researchers would not have. Contact information for data requests are available through the FDOH website: https://www.floridahealth.gov/diseases-and-conditions/mosquito-borne-diseases/surveillance.html. All data preparation and analyses were performed in R using existing functions available through tidyverse and the landscapemetrics, party, and glmmTMB packages.
Funding: Lindsay P. Campbell is supported by the USDA National Institute of Food and Agriculture, Hatch project number 1021482; Yasmin Tavares is supported through the Columbia University Fellowship, and this work was partially supported through the University of Florida IFAS Early Career Research Fund and the University of Florida Research Opportunity Seed Fund.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Landscape composition and configuration can have direct impacts on the distribution of zoonotic mosquito-borne disease systems [1]. Myriad factors, including mosquito density [2], air temperature [3], precipitation [4], humidity [5], human population distribution [6], and assemblages of hosts [7], are all elements that drive the transmission of vector-borne pathogens. However, these variables either shape the landscape (i.e., human population distribution, air temperature, precipitation) or are shaped by the landscape (i.e., mosquito density and assemblages of hosts species) [8, 9]. Thus, landscape composition and configuration serve as a proxy for geological, anthropological and climatological processes, as well as the biological communities that inhabit it. Quantifying how landscape composition and configuration impact arbovirus transmission ecology can provide a needed basis for predicting risk in a geographic area [10–13]. However, one challenge is that regional variation in the distributions of arthropod vectors and vertebrate hosts means either or both can occupy a range of differing habitats, particularly in multi-vector and multi-host systems [1]. The result is that identifying consistent and generalizable landscape drivers of transmission hazard can be challenging [14, 15], particularly across large geographic extents [16, 17].
West Nile virus (family Flaviviridae, genus Flavivirus) is one such multi-host and multi-vector zoonotic arbovirus system maintained between mosquito vectors (primarily in the genus Culex) and avian hosts (mainly in the Passeriformes and Columbiformes orders) [14, 18]. The virus has the broadest geographic distribution of any arbovirus, and it continues to expand in geographic range [19]. WNV is the leading cause of mosquito-borne human arbovirus disease in the U.S. and has a 10% mortality rate among infected people who develop neuroinvasive disease [20]. After its arrival to North America in 1999, substantial resources have been invested to understand the ecology of the WNV system in the U.S., including multiple studies examining landscape correlates with transmission to identify where risks may be greatest [18, 21]. However, landscape studies of WNV-positive mosquito pools or distributions of human cases have revealed differing results.
For example, in Dallas, TX urbanized areas were associated with WNV positive mosquito pools during a 2012 human outbreak where Culex quinquefasciatus Say (Diptera: Culicidae) mosquitoes are abundant [22]. In Chicago, IL and Detroit, MI, high incidence of WN disease in humans occurred in more suburban neighborhoods, areas with moderate vegetation, and moderate population density where Culex pipiens Linnaeus (Diptera: Culicidae) is often abundant [23, 24] found increased human disease risk with more fragmented habitats in Suffolk County, NY. In the southeastern U.S., urban and semi-urban landscapes have been linked to elevated WNV seroprevalence in birds in Georgia with a strong association with human housing density [21, 25, 26]. Similar results have been found in Southern California where developed landscapes and neglected swimming pools providing habitat for Cx. quinquefasciatus mosquitoes were associated with greater human incidence [27–29]. Conversely, population density had a negative correlation with WN incidence in Louisiana where Culex nigripalpus Theobald may play a role in transmission [30].
In other regions, agricultural land may increase transmission risk compared to adjacent areas. In California, transmission patterns were linked with fast-growing rice fields or more generic cropland preferred by Culex tarsalis Coquillett (Diptera: Culicidae) in the Central Valley [31]. In addition, irrigated agricultural landscapes and rural landscapes, which are the preferred habitat of Cx. tarsalis, were correlated with WNV transmission in the Pacific Northwest, Iowa, and Colorado [17, 32–34], while in South Dakota, irrigated agriculture was not correlated with human incidence [35].
In Florida, USA, little is known about landscape correlations and WNV. Florida encompasses < 1% of the WNV total reported human cases in the U.S. [36]. However, monitoring of virus activity through an ongoing sentinel chicken program conducted by the Florida Department of Health (FDOH) in conjunction with mosquito control programs indicates that the virus is endemic and transmission occurs statewide [37]. Multiple mosquito vectors and avian hosts are distributed throughout Florida. Culex nigripalpus and Cx. quinquefasciatus are considered the main mosquito vectors [15, 38], however additional species i.e. Culex coronator (Dyar & Knab) (Diptera: Culicidae) and Culex erraticus (Dyar & Knab) (Diptera: Culicidae) may also contribute to maintenance and/or spillover of the virus in the natural environment [39, 40]. These species occupy a range of urban and rural habitats, using artificial containers, roadside ditches, and agricultural ditches, among others as breeding sites [41, 42]. Culex nigripalpus and Cx. quinquefasciatus feed upon a wide variety of vertebrate hosts in Florida, including birds, mammals, and reptiles [43–45]. In addition, multiple Passeriformes birds that can serve as WNV hosts are found across a broad range of landscapes in Florida for all or a portion of the year, including American Robin (Turdus migratorius), Common Grackle (Quiscalus quiscula) [46] and Red-Winged Blackbird (Agelaius phoeniceus).
Despite consistent protocols for field sampling, standardized testing, and the wide distribution of the FDOH sentinel chicken program across the state, few studies have leveraged this resource to better understand the landscape ecology of WNV in Florida. Previous landscape studies focused primarily on land cover composition to predict the potential distribution of different components of the WNV system using ecological niche models [47, 48], but did not consider land cover configuration. In addition, although [49] considered edge density of multiple land cover types when analyzing 2018 WNV sentinel chicken seroconversion in northeastern Florida, no land cover variables were informative, demonstrating the need for additional analyses across multiple regions in the state.
Here, we leverage 2018 Florida Sentinel Chicken Program (FSCP) data from individual coop sites to investigate the landscape ecology of WNV sentinel chicken seroconversion in Florida. We examine landscape composition and configuration across multiple land cover types at a statewide scale and within individual bioclimatic regions in Florida. These land cover categories represent a variety of habitats for Cx. nigripalpus and Cx. quinquefasciatus mosquito densities and avian host assemblages, human population density (i.e. low, medium, and high intensity developed land cover [50], and microclimatic conditions [51, 52] (i.e. cooler shaded areas with higher relative humidity under forested canopies vs. warmer temperatures and lower relative humidity in adjacent agricultural plots receiving direct sunlight). Examination of edge habitats are representative of additional opportunities for host/vector interactions [53–55] and potential distributions of aquatic habitats important to mosquito development following precipitation events (i.e. roadside ditches) [56].
We focus on seroconversion rates for 2018, which had the greatest number of positive chickens in 10 years and were distributed broadly across the state [57]. Although little is known about specific landscapes associated with WNV seroconversion in Florida, we expect that not only the composition but the configuration of landscapes surrounding sentinel chicken coops will be important predictors of positive WNV seroconversion, including edge habitats that may affect the distribution of mosquitoes and opportunities for host interactions [54].
Overall, we expect that greater percentages of more fragmented suburban or semi-rural landscapes that support both Cx. quinquefasciatus and Cx. nigripalpus mosquito vectors and multiple WNV avian hosts will be associated with higher sentinel chicken seroconversion. However, we also expect to find regional differences in the importance of specific land cover classes to WNV seroconversion, e.g. greater agricultural land composition in central Florida will have a positive correlation with WNV seroconversion. Because the scale at which landscapes affect WNV seroconversion is unknown, we examine landscape composition and configuration summarized across multiple distances from sentinel chicken coops. Understanding these associations has the potential to reveal needed information about WNV transmission ecology in this region, while contributing to a broader understanding of multi-vector/multi-host vector-borne disease distributions across heterogeneous landscapes.
Materials and methods
The state of Florida is predominantly peninsular and located in the southeastern portion of the United States. Florida climate is characterized as humid subtropical in the northern and central regions and as equatorial monsoonal with dry winters in the southern region of the state. In 2018, annual maximum temperatures ranged from 40°C to 42°C and annual minimum temperatures -7°C to -5°C, with 1,540 mm statewide annual average precipitation in 2018 [58]. Florida landscapes are diverse (Fig 1) including woody wetlands, forested areas, and pastures predominantly found in the northwest and north regions, while a mix of herbaceous wetlands, herbaceous agricultural and low to medium-intensity developed landscapes shape the north- and south-central regions. In the southern portion of the state, medium to high-intensity developed land cover is found along coastal areas, and cultivated crops, woody, and herbaceous wetlands dominate inland areas (Fig 1).
Data resources used
Sentinel chicken monitoring data.
Florida Department of Health 2018 WNV sentinel chicken seroconversion data included 187 georeferenced coop locations surveyed over a 36-week time period between April 16th and December 23rd. Blood samples from chickens are screened for flavivirus antibodies using a Hemagglutination Inhibition test, and antibody-positive birds are then tested again using an IgM enzyme-linked immunosorbent assay (ELISA) to determine whether the chickens have antibodies for WNV [37]. Weekly values at each coop were assigned a value of 0 if no chickens tested WNV seropositive and a value of 1 if at least 1 chicken tested WNV seropositive. Weekly presence/absence data was then converted to the proportion of positive weeks over the 36-week sampling period for a single value at each coop, which served as the response variable in subsequent analyses. Coop locations were then designated into five climatic regions delineated by the U.S. Climate Divisions outlined by the National Oceanic and Atmospheric Administration (NOAA) (Fig 1) [59].
Land cover data. United States Geological Survey 2019 National Land Cover Data was downloaded from The Multi-Resolution Land Characteristics (MRLC) consortium [50], which is available at a 30m spatial resolution. Buffers surrounding coop locations were generated at 0.5 km, 2 km, 3.5 km, 5 km, and 6.5 km, and the percentage of each land cover class and the edge density of each land cover class was calculated within the buffers. Percent land cover represented the overall composition of the land cover and edge density represented the amount of edge habitat and general fragmentation surrounding each coop site.
Data processing and analysis
Variable reduction.
Landscape metrics at coop sites within each region were then binned into four general classes: forest, developed, wetland, and cropland. For example, the general wetland class included landscape metrics at all buffer distances for herbaceous wetland, woody wetland, and open water. A table showing the binned classes is available in S1 Table. A conditional random forest was then run with all landscape metrics for each class and distance within the general class for the purpose of variable reduction. Conditional random forests were run using the ‘party’ package in R with ‘mtry’ values equal to the square root of the number of variables in the model and iterations ranging from 3,000 to 30,000 based on the stability of model results [50–52, 60–62]. Variable importance values were generated for each general land cover category within each region, and the variable with the greatest importance in the random forest output was selected as the environmental covariate in subsequent model runs. The result was a total of four landscape variables for each region and the state representing general forest, developed, wetland, and cropland land cover categories to be included as predictor variables in model runs. We then calculated variance inflation factor (VIF) values for environmental variables for each landscape candidate model as a test for multicollinearity.
Model runs and spatial autocorrelation.
Binomial generalized linear mixed effects models (GLMMs) with the proportion of seropositive coops, weighted by the number of sampling weeks, served as the response variable in the regional landscape models, and the forest, developed, wetland, and cropland variables identified in the conditional random forest model served as the predictor variables in each regional model. Models included a site-level random effect to reduce overdispersion and were run using the ‘glmmTMB’ package in R and the ‘dredge’ function in the ‘MuMIn’ package was used to generate models for all combinations of variables in each candidate set [63, 64]. Models were ranked from lowest to highest Akaike’s Information Criterion (AIC) scores, AIC weights were calculated, and we used a delta AIC value of > 2 as a threshold to identify the “best” set of models [65, 66]. Residual spatial autocorrelation was investigated using the ‘Moran.I’ function in the package ‘ape’ [67]. If significant spatial autocorrelation was present, a spatial random effect term with an exponential distance decay function available in ‘glmmTMB’ was added and the candidate set was rerun [64]. Additional model diagnostics checking for non-parametric dispersion were performed in the R package ‘DHARMa’ [68]. Effect plots for variables in the top-ranking model for each region were generated using the ‘ggpredict’ function in the ‘ggeffects’ package in R [69].
Results
Results from conditional random forests indicated variation in the importance of landscape variables across regions. Variables included in the four general land cover classes for regional candidate sets and the overall statewide model included both percent land cover and edge density variables, but varied in buffer distances and specific land cover class (S1 Table).
As expected, results from GLMMs indicated that the importance of landscape correlations with WNV seroconversion varied across regions. However, in the statewide model and the northern most region, we did not find strong support from any of the variables included in the models, which was indicated by the intercept-only model ranking as the “best” model (Table 1). We also found weak support for correlations with edge density of forest within 2 km in the northwest region (p--value = 0.391) (Fig 2). However, we found strong support for landscape composition and configuration with WNV seroconversion in the remaining regions (all other p--values < = 0.05).
Positive values indicate greater percentages or edge density and negative values indicate lower percentages or edge density values.
In the north-central region, we found a non-linear and negative effect of percentage low intensity developed land cover within 5 km on WNV seroconversion, meaning that coops surrounded by greater percentages of more rural but populated areas had lower seroconversion rates (Fig 2). Three models comprised the best set of models in the north-central region, and percentage low intensity developed land cover within 5 km was included in each model. The most parsimonious model had strong support with an AICw of 0.457 (S2 Table).
In the south-central region, we found a positive effect of edge density of forest within 0.5 km of coop sites and a negative effect of edge density of woody wetlands within 2 km on WNV seroconversion, indicating that more fragmented forests at near distances and more intact woody wetlands up to 2 km were correlated with WNV transmission (Fig 2). Four models comprised the best set of models in this region, and edge density of woody wetlands within 2 km was included in all four models, while edge density of forest was included in three models. The most parsimonious model had moderate support with an AICw of 0.227 (S2 Table).
In the south region, we found that edge density of open water within 5 km of coop sites had a negative effect on WNV seroconversion, meaning that seroconversion decreased when ocean or large bodies of water were located within 5 km of coop sites. In this region, 7 models comprised the best set of models and the most parsimonious model had the lowest support of all the regions with an AICw of 0.16. Edge density of open water was included in five of the seven best models (S2 Table).
Residual diagnostics for each of the best-performing models indicated that in the northwest and north-central regions, some non-parametric dispersion was present. Moran’s I tests for residual spatial autocorrelation indicated that all regions except for the south region required the inclusion of a spatial random effect. A summary of the best set of models for each region and the full model results are available in S2 and S3 Tables.
Discussion
The sentinel chicken program was developed with the goal of monitoring mosquito-borne pathogen transmission in the environment to inform potential risk. Our study is the first to investigate landscape composition and configuration associations with WNV seropositive sentinel chickens across Florida, and within different bioclimatic regions. A key result is that we found no consistent statewide landscape predictors for WNV seroconversion but found clear within-state regional associations in three out of five study areas. These regional associations include significant landscape composition and configuration within near distances (0.5 km) and broader distances (2 km) from sentinel chicken coops. These results suggest that in Florida, broader extent studies may dampen relevant landscape signals present at more local to regional scales when examining landscapes alone. This work further demonstrates the many challenges with understanding and predicting zoonotic arbovirus risk in complex, human dominated systems but also provides insights about the landscape ecology of WNV in Florida and some key next steps, which we cover below.
When examining individual bioclimatic regions, we found that landscape configuration was particularly informative in the south-central region of Florida, where more fragmented forested areas near chicken coops (0.5 km) and intact woody wetlands at further distances (2 km) were positively correlated with WNV seroconversion. Within this region, chicken coops were distributed primarily in or near coastal communities within developed landscapes along the Atlantic and Gulf Coasts (Fig 1). Fragmented forested areas in this study region may be due to roadways, which are often lined with ditches for water management that support Cx. nigripalpus development [56]. In addition, these patterns can arise from dense but fragmented tree coverage within residential areas where Cx. quinquefasciatus thrives [70], and several WNV amplifying hosts are widely found in fragmented areas across these landscapes. Similar to other geographic regions, Florida is undergoing rapid and transformational land cover change with intensifying development that can result in fragmented habitats [71]. Continued investigation of mosquito vector and avian host interactions across fragmented forested habitats in the south-central region has the potential to provide additional insight into WNV ecology and transmission hazard in this area.
We also found in the south-central region that more aggregated woody wetland areas within 2 km of chicken coops were an important predictor of WNV positive seroconversion (Fig 2). This result contradicted previous findings in the northeastern U.S. and in the Chicago Metropolitan Area that found a negative correlation of percentages of woody wetland and herbaceous wetland areas with WNV transmission [16, 49, 72]. However, in addition to differences in the vegetative species comprising woody wetlands in these temperate regions, these studies focused on percentage land cover but did not examine configuration. In the south-central region of our study area, more aggregated patches of woody wetland are often associated with landscapes further from the coast, which could also be an indicator of a transition from populated areas to more semi-rural to rural areas. However, when considering distributions of WNV vector species in this region [73], demonstrated that in Manatee County, a coastal county within the south-central region, land use predicts the greatest proportion of WNV vector-competent mosquitoes in urbanized areas near the coast, and [74] found no correlation between woody wetlands and Cx. quinquefasciatus mosquitoes in New Orleans, Louisiana. Empirical field studies may yield valuable insight into how particular aspects of woody wetlands, including their configuration across the landscape, support vector-host interactions and elevate WNV distributions in this area.
The importance of landscape composition also varied regionally. In the north-central region (Fig 2), we found a non-linear effect of percentage of low-intensity developed land cover within 5 km of chicken coops. Low-intensity developed land cover includes single-family housing units mixed with vegetative cover and impervious surfaces, often in transition areas to more rural landscapes [50]. These results indicate that very high percentages of rural developed landscapes and areas that include no rural developed landscapes have low seroconversion, but that a mid-level percentage of these landscapes are a significant predictor of positive seroconversion. Overall, these results supported our expectation that more semi-urban areas would be positively correlated with WNV seroconversion. However, consistent with several other landscape studies in the U.S. [16, 21], this finding was constrained to a relatively small geographic area and similar results were not found in additional study regions in Florida. Despite the relatively small geographic area, this information has the potential to be useful to monitoring and control in this region. Additional sampling, including mosquito pool testing and blood meal analyses, will provide further opportunities to understand WNV transmission ecology within semi-rural areas in this region.
Although agricultural landscapes were correlated with WNV in multiple regions in the U.S. [17, 32–34], we did not find evidence that agricultural landscapes were correlated with WNV seroconversion at the state level or in our regional analyses in Florida, even though these habitats can support high abundances of Cx. nigripalpus and Cx. quinquefasciatus mosquitoes [41, 42]. This finding, and our overall finding of no statewide predictors of WNV seroconversion demonstrates further the challenges to generalizing results from landscape studies of the WNV system across geographic areas. The varied and nuanced relationships observed between WNV seroconversion and landscape composition and configuration in this study may be a result of the differences in the ecologies of the two putative vector species, Cx. nigripalpus and Cx. quinquefasciatus in Florida. [44] demonstrated that female numbers and class-level host use of Cx. nigripalpus and Cx. quinquefasciatus changes on an urban-to-rural gradient, but in different ways. For example, while numbers of blood-engorged Cx. nigripalpus females were relatively consistent across urban, suburban and rural sites (range 140–191), numbers of blood-engorged Cx. quinquefasciatus decreased precipitously across the same sites (386 in urban, 48 in suburban, 7 in rural). For Cx. nigripalpus the fraction of blood meals from avian hosts increased from urban to rural gradient (38.2% in urban, 54.1% in suburban, 82.1% in rural) but was relatively consistent across the urban rural gradient for Cx. quinquefasciatus (range 20.7–28.6%) [44]. These differences in the distributions and host use behaviors of the two vector species, along with variation in avian host species distributions and competency highlight challenges to disentangling landscape correlations with multi-vector/multi-host systems such as WNV.
In addition to variation in host distributions and host use among mosquito vector species, seasonal variation and heterogeneity in vector densities driven by abiotic factors, including temperature, precipitation, and humidity can affect the distribution of WNV transmission [2, 5]. These factors can also affect the timing and distribution of migration for some avian host species, and changes in avian host communities due to migratory dynamics can impact interactions between susceptible hosts and mosquito vectors, particularly across fragmented landscapes [7]. Future studies incorporating seasonal factors may provide additional information toward understanding WNV transmission ecology in Florida. While our study investigated landscape correlations with WNV transmission to sentinel chickens across a robust set of surveillance sites in Florida, some limitations exist. Here, we examined patterns resulting from a single elevated seropositive year. However, additional variation may be detected across longer time periods. Finally, we examined land cover data specific to the 2018 time period, but lagged effects of anthropogenic land cover change and ongoing landscape dynamics may also provide new insight into the ecology of WNV in Florida.
Despite these limitations, results from this study provide new information about opportunities and challenges to understanding the landscape ecology of WNV across the state. Here, we can conclude that it is beneficial to investigate the landscapes using a regional approach and that both landscape composition and configuration can be informative predictors of WNV within specific regions, when focusing on sentinel chicken seroconversion as an indicator of virus activity. Leveraging longer-term interannual time series of FDOH WNV sentinel chicken surveillance data to examine dynamic climate and landscape factors together may reveal stronger predictors of seroconversion. In addition, the inclusion of targeted field studies is needed to disentangle host and vector dynamics across these environments. This combined approach will be critical to understanding and predicting WNV transmission ecology in this region.
Supporting information
S1 Table. Variables demonstrating the greatest importance from each random forest run for developed, wetland, forest, and cropland categories in each bioclimatic region in Florida; “PL” indicates percent land cover and “ED” indicates edge density metrics.
Pearson’s coefficient (r) values for each variable with WNV seroconversion rate weighted by number of weeks sampled.
https://doi.org/10.1371/journal.pone.0305510.s001
(XLSX)
S2 Table. R glmmTMB results of the “best” set of models for each bioclimatic region in Florida; “pl” indicates percent land cover and “ed” indicates edge density metrics.
https://doi.org/10.1371/journal.pone.0305510.s002
(XLSX)
S3 Table. Values for Moran I for spatial autocorrelation calculations.
https://doi.org/10.1371/journal.pone.0305510.s003
(XLSX)
Acknowledgments
We would like to acknowledge the hard work of the Florida Department of Health Arbovirus Surveillance program and individual Florida mosquito control programs who maintain, test, and distribute results of the sentinel chicken program and provided data for these analyses. We would also like to thank Amely M. Bauer for her help with data visualization and figure design.
References
- 1. Ziegler M. Landscapes of disease. Landscapes. 2016;17: 99–107.
- 2. Li Y, Zhou G, Zhong S, Wang X, Zhong D, Hemming-Schroeder E, et al. Spatial heterogeneity and temporal dynamics of mosquito population density and community structure in Hainan Island, China. Parasit Vectors. 2020;13: 444. pmid:32887654
- 3. Mordecai EA, Caldwell JM, Grossman MK, Lippi CA, Johnson LR, Neira M, et al. Thermal biology of mosquito-borne disease. Ecol Lett. 2019;22: 1690–1708. pmid:31286630
- 4. Gage KL, Burkot TR, Eisen RJ, Hayes EB. Climate and vectorborne diseases. Am J Prev Med. 2008;35: 436–450. pmid:18929970
- 5. Brown JJ, Pascual M, Wimberly MC, Johnson LR, Murdock CC. Humidity—The overlooked variable in the thermal biology of mosquito-borne disease. Ecol Lett. 2023;26: 1029–1049. pmid:37349261
- 6. de Souza WM, Weaver SC. Effects of climate change and human activities on vector-borne diseases. Nat Rev Microbiol. 2024;22: 476–491. pmid:38486116
- 7. Krol L, Remmerswaal L, Groen M, van der Beek JG, Sikkema RS, Dellar M, et al. Landscape level associations between birds, mosquitoes and microclimates: Possible consequences for disease transmission? Parasit Vectors. 2024;17: 156. pmid:38532512
- 8. Wilcox BA, Gubler DJ. Disease ecology and the global emergence of zoonotic pathogens. Environ Health Prev Med. 2005;10: 263–272. pmid:21432130
- 9.
Ostfeld RS, Keesing F, Eviner VT. Infectious disease ecology: Effects of ecosystems on disease and of disease on ecosystems. Ostfeld RS, Keesing F, Eviner VT, editors. Princeton, NJ: Princeton University Press; 2008. https://doi.org/10.1515/9781400837885
- 10.
Pavlovsky EN. The natural nidus of a disease as a pathobiocenose. The Natural Nidality of Transmissible Disease Univ.
- 11. Lambin EF, Tran A, Vanwambeke SO, Linard C, Soti V. Pathogenic landscapes: Interactions between land, people, disease vectors, and their animal hosts. Int J Health Geogr. 2010;9: 54. pmid:20979609
- 12. Kilpatrick AM. Globalization, land use, and the invasion of West Nile virus. Science. 2011;334: 323–327. pmid:22021850
- 13. Reisen WK. Landscape epidemiology of vector-borne diseases. Annu Rev Entomol. 2010;55: 461–483. pmid:19737082
- 14. Colpitts TM, Conway MJ, Montgomery RR, Fikrig E. West Nile Virus: Biology, transmission, and human infection. Clin Microbiol Rev. 2012;25: 635–648. pmid:23034323
- 15. Rochlin I, Faraji A, Healy K, Andreadis TG. West Nile virus mosquito vectors in North America. J Med Entomol. 2019;56: 1475–1490. pmid:31549725
- 16. Bowden SE, Magori K, Drake JM. Regional differences in the association between land cover and West Nile virus disease incidence in humans in the United States. Am J Trop Med Hyg. 2011;84: 234–238. pmid:21292890
- 17. DeGroote JP, Sugumaran R, Brend SM, Tucker BJ, Bartholomay LC. Landscape, demographic, entomological, and climatic associations with human disease incidence of West Nile virus in the state of Iowa, USA. Int J Health Geogr. 2008;7: 19. pmid:18452604
- 18. Reisen WK. Ecology of West Nile virus in North America. Viruses. 2013;5: 2079–2105. pmid:24008376
- 19. de Freitas Costa E, Streng K, Avelino de Souza Santos M, Counotte MJ. The effect of temperature on the boundary conditions of West Nile virus circulation in Europe. PLoS Negl Trop Dis. 2024;18: e0012162. pmid:38709836
- 20. Ciota AT. West Nile virus and its vectors. Curr Opin Insect Sci. 2017;22: 28–36. pmid:28805636
- 21. DeGroote JP, Sugumaran R. National and regional associations between human West Nile virus incidence and demographic, landscape, and land use conditions in the coterminous United States. Vector Borne Zoonotic Dis. 2012;12: 657–665. pmid:22607071
- 22. Poh KC, Medeiros MCI, Hamer GL. Landscape and demographic determinants of Culex infection with West Nile virus during the 2012 epidemic in Dallas County, TX. Spat Spatiotemporal Epidemiol. 2020;33: 100336.
- 23. Ruiz MO, Walker ED, Foster ES, Haramis LD, Kitron UD. Association of West Nile virus illness and urban landscapes in Chicago and Detroit. Int J Health Geogr. 2007;6: 10. pmid:17352825
- 24. Rochlin I, Turbow D, Gomez F, Ninivaggi DV, Campbell SR. Predictive mapping of human risk for West Nile virus (WNV) based on environmental and socioeconomic factors. PLoS One. 2011;6: e23280. pmid:21853103
- 25. Gibbs SEJ, Wimberly MC, Madden M, Masour J, Yabsley MJ, Stallknecht DE. Factors affecting the geographic distribution of West Nile virus in Georgia, USA: 2002–2004. Vector Borne Zoonotic Dis. 2006;6: 73–82. pmid:16584329
- 26. Levine RS, Mead DG, Kitron UD. Limited spillover to humans from West Nile Virus viremic birds in Atlanta, Georgia. Vector Borne Zoonotic Dis. 2013;13: 812–817. pmid:24107200
- 27. Reisen WK, Takahashi RM, Carroll BD, Quiring R. Delinquent mortgages, neglected swimming pools, and West Nile virus, California. Emerg Infect Dis. 2009;15: 508b–5509.
- 28. Kwan JL, Kluh S, Madon MB, Nguyen DV, Barker CM, Reisen WK. Sentinel chicken seroconversions track tangential transmission of West Nile virus to humans in the greater Los Angeles area of California. Am J Trop Med Hyg. 2010;83: 1137–1145. pmid:21036853
- 29. Snyder RE, Feiszli T, Foss L, Messenger S, Fang Y, Barker CM, et al. West Nile virus in California, 2003–2018: A persistent threat. PLoS Negl Trop Dis. 2020;14: e0008841. pmid:33206634
- 30. Ezenwa VO, Milheim LE, Coffey MF, Godsey MS, King RJ, Guptill SC. Land cover variation and West Nile virus prevalence: Patterns, processes, and implications for disease control. Vector Borne Zoonotic Dis. 2007;7: 173–180. pmid:17627435
- 31. Kovach TJ, Kilpatrick AM. Increased Human Incidence of West Nile Virus disease near rice fields in California but not in southern United States. Am J Trop Med Hyg. 2018;99: 222–228. pmid:29714160
- 32. Eisen L, Barker CM, Moore CG, Pape WJ, Winters AM, Cheronis N. Irrigated agriculture is an important risk factor for West Nile virus disease in the hyperendemic Larimer-Boulder-Weld area of north central Colorado. J Med Entomol. 2010;47: 939–951. pmid:20939393
- 33. Crowder DW, Dykstra EA, Brauner JM, Duffy A, Reed C, Martin E, et al. West nile virus prevalence across landscapes is mediated by local effects of agriculture on vector and host communities. PLoS One. 2013;8: e55006. pmid:23383032
- 34. Dunphy BM, Kovach KB, Gehrke EJ, Field EN, Rowley WA, Bartholomay LC, et al. Long-term surveillance defines spatial and temporal patterns implicating Culex tarsalis as the primary vector of West Nile virus. Sci Rep. 2019;9: 6637.
- 35. Chuang T-W, Hockett CW, Kightlinger L, Wimberly MC. Landscape-level spatial patterns of West Nile virus risk in the northern Great Plains. Am J Trop Med Hyg. 2012;86: 724–731. pmid:22492161
- 36.
CDC. Centers for Disease Control and Prevention. West Nile Virus Disease. 2021 [cited 23 Nov 2022]. Available: https://www.cdc.gov/westnile/index.html
- 37.
FDOH. Florida Department of Health. Non-Human Mosquito-Borne Disease Monitoring Activities. 2021 [cited 20 Nov 2021]. Available: http://www.floridahealth.gov/diseases-and-conditions/mosquito-borne-diseases/_documents/guidebook-chapter-ten.pdf
- 38. Vitek CJ, Richards SL, Mores CN, Day JF, Lord CC. Arbovirus transmission by Culex nigripalpus in Florida, 2005. J Med Entomol. 2008;45: 483–493.
- 39. Burkett-Cadena ND, Hassan HK, Eubanks MD, Cupp EW, Unnasch TR. Winter severity predicts the timing of host shifts in the mosquito Culex erraticus. Biol Lett. 2012;8: 567–569.
- 40. Alto BW, Connelly CR, O’Meara GF, Hickman D, Karr N. Reproductive biology and susceptibility of Florida Culex coronator to infection with West Nile virus. Vector Borne Zoonotic Dis. 2014;14: 606–614.
- 41. Day JF, Tabachnick WJ, Smartt CT. Factors that influence the transmission of West Nile virus in Florida. J Med Entomol. 2015;52: 743–754. pmid:26336216
- 42.
Wilkerson RC, Linton Y-M, Strickman D. Mosquitoes of the World. Baltimore, MD: Johns Hopkins University Press; 2021.
- 43. Edman JD. Host-feeding patterns of Florida mosquitoes: III. Culex (Culex) and Culex (Neoculex)1. J Med Entomol. 1974;11: 95–104.
- 44. Hancock C, Camp JV. Habitat-specific host selection patterns of Culex quinquefasciatus and Culex nigripalpus in Florida. J Am Mosq Control Assoc. 2022;38: 83–91.
- 45. Reeves LE, Burkett-Cadena ND. Lizards are important hosts for zoonotic Flavivirus vectors, subgenus Culex, in the southern USA. Frontiers in Tropical Diseases. 2022;3.
- 46. Hostetler M, Archer J-M. Building for birds evaluation tool: Breeding and wintering habitat for forest birds. EDIS. 2017;2017: 17.
- 47. Sallam MF, Xue R-D, Pereira RM, Koehler PG. Ecological niche modeling of mosquito vectors of West Nile virus in St. John’s County, Florida, USA. Parasit Vectors. 2016;9: 371. pmid:27357295
- 48. Beeman SP, Morrison AM, Unnasch TR, Unnasch RS. Ensemble ecological niche modeling of West Nile virus probability in Florida. PLoS One. 2021;16: e0256868. pmid:34624026
- 49. Myer MH, Campbell SR, Johnston JM. Spatiotemporal modeling of ecological and sociological predictors of West Nile virus in Suffolk County, NY, mosquitoes. Ecosphere. 2017;8: e01854. pmid:30147987
- 50.
Dewitz, J., and U.S. Geological Survey. National Land Cover Database (NLCD) 2019 Products (ver. 2. 0, June 2021): U.S. Geological Survey data release (ver. 2. 0, June 2021): U.S. Geological Survey data release.
- 51.
Zellweger F. Florian et al. Trends in Ecology & Evolution. 34: 327–341.
- 52. De Frenne P, Lenoir J, Luoto M, Scheffers BR, Zellweger F, Aalto J, et al. Forest microclimates and climate change: Importance, drivers and future research agenda. Glob Chang Biol. 2021;27: 2279–2297. pmid:33725415
- 53. Rizzoli A, Tagliapietra V, Cagnacci F, Marini G, Arnoldi D, Rosso F, et al. Parasites and wildlife in a changing world: The vector-host- pathogen interaction as a learning case. Int J Parasitol Parasites Wildl. 2019;9: 394–401. pmid:31341772
- 54. Reiskind MH. Mosquitoes of field and forest: The scale of habitat segregation in a diverse mosquito assemblage. Medical and Veterinary Entomology. 2017;31: 44–54. pmid:27759165
- 55. Gottwalt A. Impacts of Deforestation on Vector-borne Disease Incidence. CU JGH. 2013;3: 16–19.
- 56.
Day JF. The Florida SLE Mosquito, Culex (Culex) nigripalpus Theobald (Insecta: Diptera: Culicidae). University of Florida, Vero Beach, FL. 2004. Available: https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=cc83c8db3dc281311c931679776cedf339b01532
- 57.
Florida Arbovirus Surveillance. [cited 31 May 2024]. Available: https://www.floridahealth.gov/diseases-and-conditions/mosquito-borne-diseases/_documents/2018-week-52-arbovirus-surveillance-report.pdf
- 58. Brouillette DJ. Florida Annual Climate Summary for 2018. Florida Climate Center.
- 59.
History of the U.S. Climate Divisional Dataset. National Oceanic and Atmospheric Administration. 2023. Available: https://www.ncei.noaa.gov/access/monitoring/dyk/us-climate-divisions
- 60. Hothorn T, Bühlmann P, Dudoit S, Molinaro A, van der Laan MJ. Survival ensembles. Biostatistics. 2006;7: 355–373. pmid:16344280
- 61. Strobl C, Boulesteix A-L, Kneib T, Augustin T, Zeileis A. Conditional variable importance for random forests. BMC Bioinformatics. 2008;9: 307. pmid:18620558
- 62. Strobl C, Boulesteix A-L, Zeileis A, Hothorn T. Bias in random forest variable importance measures: illustrations, sources and a solution. BMC Bioinformatics. 2007;8: 25. pmid:17254353
- 63.
Bartoń K. MuMIn: MultiModel Inference. R package version 1475. 2023. Available: https://CRAN.R-project.org/package=MuMIn
- 64. Brooks M, Kristensen K, van Benthem K, Magnusson A, Berg C, Nielsen A, et al. GlmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. R J. 2017;9: 378.
- 65. Akaike H. Likelihood of a model and information criteria. J Econom. 1981;16: 3–14.
- 66.
Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. Springer Science & Business Media; 2007.
- 67. Paradis E, Schliep K. ape 5.0: An environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics. 2019;35: 526–528. pmid:30016406
- 68.
Hartig F. DHARMa: Residual Diagnostics for Hierarchical (Multi-Level/Mixed) Regression Models, R Package Version 0.4. 5. 2022. 2022.
- 69. Lüdecke D. Ggeffects: Tidy data frames of marginal effects from regression models. J Open Source Softw. 2018;3: 772.
- 70. Wilke ABB, Vasquez C, Carvajal A, Moreno M, Fuller DO, Cardenas G, et al. Urbanization favors the proliferation of Aedes aegypti and Culex quinquefasciatus in urban areas of Miami-Dade County, Florida. Sci Rep. 2021;11: 22989. pmid:34836970
- 71.
Volk M, Hoctor T, Nettles B, Hilsenbeck R, Putz F. Florida land use and land cover change in the past 100 years. Florida’s Climate: Changes, Variations, & Impacts. Florida Climate Institute; 2017.
- 72. Johnson BJ, Munafo K, Shappell L, Tsipoura N, Robson M, Ehrenfeld J, et al. The roles of mosquito and bird communities on the prevalence of West Nile virus in urban wetland and residential habitats. Urban Ecosyst. 2012;15: 513–531. pmid:25484570
- 73. Bauer AM, Guralnick RP, Whitehead SA, Barve N, Allen JM, Campbell LP. Land use predicts proportion of West Nile virus vector‐competent mosquitoes. Ecosphere. 2024;15.
- 74. Moise IK, Riegel C, Muturi EJ. Environmental and social-demographic predictors of the southern house mosquito Culex quinquefasciatus in New Orleans, Louisiana. Parasit Vectors. 2018;11. pmid:29665862