https://orcid.org/0000-0001-8082-6352
Figures
Abstract
Risk of spillover and spillback of mosquito-borne viruses in the neotropics, including yellow fever, dengue, Zika (Flaviviridae: Flavivirus), chikungunya, and Mayaro (Togaviridae: Alphavirus) viruses, is highest at ecotones where humans, monkeys, and mosquitoes coexist. With a view to identifying potential bridge vectors, we investigated changes in mosquito community composition and environmental variables at ground level at distances of 0, 500, 1000, and 2000 m from the edge of a rainforest reserve bordering the city of Manaus in the central Brazilian Amazon. During two rainy seasons in 2019 and 2020, we sampled 9,467 mosquitoes at 244 unique sites using BG-Sentinel traps, hand-nets, and Prokopack aspirators. Species richness and diversity were generally higher at 0 m and 500 m than at 1000 m and 2000 m, while mosquito community composition changed considerably between the forest edge and 500 m before stabilizing by 1000 m. Shifts in environmental variables mainly occurred between the edge and 500 m, and the occurrence of key taxa (Aedes albopictus, Ae. scapularis, Limatus durhamii, Psorophora amazonica, Haemagogus, and Sabethes) was associated with one or more of these variables. Sites where Ae. aegypti and Ae. albopictus were detected had significantly higher surrounding mean NDBI (Normalized Difference Built-up Index) values than sites where they were not detected, while the opposite was true for Sabethes mosquitoes. Our findings suggest that major changes in mosquito communities and environmental variables occur within 500 m of the forest edge, where there is high risk for contact with both urban and sylvatic vectors. By 1000 m, conditions stabilize, species diversity decreases, and forest mosquitoes predominate. Environmental variables associated with the occurrence of key taxa may be leveraged to characterize suitable habitat and refine risk models for pathogen spillover and spillback.
Author summary
Mosquito-borne viruses, including yellow fever, dengue, Zika, chikungunya, and Mayaro viruses, derive from sylvatic transmission cycles involving forest mosquitoes and wild animals. Occasionally, these sylvatic viruses spill over into humans, which can lead to limited disease outbreaks or sustained human transmission by urban mosquitoes. When translocated by humans to new geographic regions, these viruses have potential to spill back into novel sylvatic cycles, creating a lasting threat to human health. Ecotones, such as forest edges, where humans, mosquitoes, and monkeys coexist, are at high risk of spillover and spillback. We investigated changes in mosquito community composition at incremental distances up to 2000 m into rainforest bordering the Amazonian city of Manaus, where the above viruses circulate in urban and sylvatic cycles. We also characterized mosquito habitat and analyzed associations between key vector species and environmental variables like temperature and vegetation. Our results show that mosquito communities and environmental variables change abruptly within 500 m of the forest edge, where there is high risk for contact of both monkeys and humans with both urban and forest mosquitoes. By 1000 m, conditions stabilize, and forest mosquitoes predominate. Environmental variables associated with the presence of particular mosquito taxa will help characterize suitable habitat for key vector species and refine spillover and spillback risk models for these viruses.
Citation: Hendy A, Hernandez-Acosta E, Valério D, Fé NF, Mendonça CR, Costa ER, et al. (2023) Where boundaries become bridges: Mosquito community composition, key vectors, and environmental associations at forest edges in the central Brazilian Amazon. PLoS Negl Trop Dis 17(4): e0011296. https://doi.org/10.1371/journal.pntd.0011296
Editor: Marc Choisy, University of Oxford, UNITED KINGDOM
Received: February 9, 2023; Accepted: April 10, 2023; Published: April 26, 2023
Copyright: © 2023 Hendy 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: The authors confirm that all data underlying the findings are fully available without restriction. All relevant data are within the paper and its Supporting Information files.
Funding: The research was funded in part by an International Collaborations in Infectious Disease Research (ICIDR) grant 1U01AI115577 awarded to NV and a Centers for Research in Emerging Infectious Diseases (CREID) 1U01AI151807 grant awarded to NV and KAH by the National Institutes of Health (NIH), and a Consejo Nacional de Ciencia y Tecnología fellowship awarded to EHA by the Government of Mexico, 709139. MVGL is a CNPq fellow. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Background
The potential for pathogen exchange between native and novel hosts is thought to be highest at ecosystem boundaries [1,2]. The risk of such exchange is particularly acute for arthropod-borne viruses (arboviruses), as all arboviruses affecting humans derive from wildlife transmission cycles [3]. Arboviruses transmitted by mosquitoes have the greatest public health impact, and the most prominent of these are transmitted in sylvatic cycles between non-human primates (monkeys) and mainly canopy-dwelling mosquitoes [4], including yellow fever (YFV), dengue (DENV), Zika (ZIKV) (all Flaviviridae: Flavivirus), and chikungunya (CHIKV) (Togaviridae: Alphavirus) [5]. Spillover of these viruses from their sylvatic cycles, mediated by mosquito bridge vectors, can lead to isolated human infections, localized epidemics, or establishment of human endemic cycles with global reach [3].
Of all the mosquito-borne viruses, the role of ecotones in spillover has been most firmly established for YFV [1]. This virus emerged from its ancestral African sylvatic cycle and was introduced to the Americas approximately 400 years ago [6], where it soon spilled back into a novel sylvatic cycle involving neotropical monkeys and Haemagogus and Sabethes species mosquitoes. In recent decades, DENV, ZIKV, and CHIKV have also emerged or resurged in the Americas where, up to now, they have been transmitted exclusively in urban cycles by domestic Aedes aegypti and peridomestic Ae. albopictus mosquitoes [3,4]. In areas where humans, monkeys, and mosquitoes coexist, these viruses could potentially spill back into novel sylvatic cycles which may be impossible to eradicate, thereby creating a lasting threat to human health [7–9].
While it is widely accepted that risk of arbovirus spillover and spillback is elevated at forest edges, the mechanisms for such effects are incompletely understood. Borremans et al. [2] posit that the “permeability” of edges for mosquito-borne arbovirus exchange will depend upon the abundance, biting rates, and competence of mosquito vectors, as well as host abundance, near edges. Moreover, edge effects on microclimate are likely to alter the ability of a given mosquito to transmit an arbovirus [10], with warmer forest edges likely enhancing vector competence. Studies investigating shifts in mosquito communities at forest edges often utilize mechanized sampling methods [11–13], delivering a biased view of mosquito diversity and abundance and no insight into biting rates. For example, we previously used BG-Sentinel traps and Prokopack aspirators to sample mosquitoes at ground level in urban forest fragments in Manaus, Brazil [13]. These aspirators are designed to sample resting, potentially blood fed mosquitoes [14], which are not necessarily anthropophilic. In contrast, BG-Sentinel traps are designed to sample host-seeking Aedes and Culex species mosquitoes [15] but are less specific than humans in attracting anthropophilic mosquitoes [16]. Consequently, the choice of sampling method skews collections towards mosquitoes attracted to associated stimuli. For instance, anthropophilic mosquitoes may be attracted to CO2 and other attractants associated with BG-Sentinel traps, but so might ornithophilic species, whereas ornithophilic species would likely not be attracted to a human wielding a hand-net.
In our earlier study of urban forest fragments [13], Ae. albopictus was the most abundant urban vector species sampled. It occurred most frequently at forest edges but penetrated at least 100 m into the forest. We have since used BG-Sentinel traps and hand-nets [17,18] to investigate vertical changes in mosquito communities at the Adolpho Ducke forest reserve bordering Manaus, showing that Haemagogus and Sabethes mosquitoes, along with Psorophora amazonica, could potentially provide routes for pathogen exchange between the forest canopy and forest floor. Our studies have shown BG-Sentinel traps to be effective for the collection of urban Aedes vectors and Psorophora species among others, while hand-net collections have yielded greater numbers of Haemagogus mosquitoes including Hg. janthinomys [13,17]. Both methods have proven reasonably effective for the collection of Sabethes species mosquitoes [17].
In this study, we investigated ground-level changes in mosquito community composition at 0, 500, 1000, and 2000 m into the forest at the same reserve, using both mechanized and human-operated sampling methods, with a view to identifying potential bridge vectors of mosquito-borne viruses. We predicted that urban species, including known Aedes vectors, would predominate at edge sites, and that considerable changes in species composition would occur between 0 m and 500 m, consistent with our previous study of interior forest parks in Manaus [13]. Beyond 500 m, we predicted that collections would be dominated by sylvatic species. Furthermore, we anticipated that the occurrence of Haemagogus and Sabethes mosquitoes would increase with increasing temperature and decreasing relative humidity, but Psorophora mosquitoes would show no such associations, in broad agreement with our previous findings [17,18].
Methods
Ethics statement
Mosquito collections were approved by local environmental authorities (SISBIO license 57003–6) and the study did not involve endangered or protected species. When sampling with hand-nets, skin was not deliberately exposed to attract mosquitoes and mosquito landing was not permitted. All collectors are listed among the co-authors or acknowledgements and were fully aware of the nature of the research. They wore trousers, a long-sleeved shirt and/or repellent to minimize the risk of being bitten and had been vaccinated against yellow fever.
Study area
The Adolpho Ducke forest reserve (Ducke) encompasses 100 km2 of primary rainforest which borders urban Manaus along the northeastern edge of the city [17,18] (Fig 1). Ducke is mainly accessed by researchers and students and is home to six monkey species: the golden-faced saki (Pithecia chrysocephala), pied tamarin (Saguinus bicolor), tufted capuchin (Sapajus apella), Guianan red howler monkey (Alouatta macconnelli), Guiana spider monkey (Ateles paniscus), and bearded saki (Chiropotes chiropotes) [19]. A grid-like network of trails arranged at 1 km intervals and covering 64 km2 allows access to much of the reserve. Manaus has undergone rapid urban expansion since the 1980s due to government incentives to develop the region, which has led to deforestation and brought people into close contact with forest edges [20]. This is especially pertinent along the southern and western edges of the reserve where dense urban areas touch the forest [19]. DENV, ZIKV, and CHIKV circulate in Manaus during the rainy season [21], which usually lasts from November to May [19]. Both YFV [22] and Mayaro virus (MAYV) (Togaviridae: Alphavirus) [23] are known to occur in rainforest near the city.
The upper left panel shows the study location in Manaus, Brazil, South America. The other panels show Manaus, the Adolpho Ducke forest reserve, and the edge between city and forest superimposed on a color infrared composite based on Landsat 8 Operational Land Imager surface reflectance imagery obtained from the USGS Earth Explorer data portal [24]. The composite was generated by placing the near infrared, red, and green spectral bands of the satellite image in the red, green, and blue color channels of the computer system. Red areas represent healthy green vegetation; cyan areas represent built-up surfaces and the sediment-laden Amazon River (Rio Solimões) which confluences with the nearly sediment-free Rio Negro at the Meeting of Waters (Encontro das Águas) south of Manaus. Country boundaries obtained from: t als://www.naturalearthdata.com/downloads/10m-cultural-vectors/10m-admin-0-countries/.
Remote sensing environmental variables
Three variables were derived from remote sensing data: land cover, Normalized Difference Vegetation Index (NDVI), and Normalized Difference Built-up Index (NDBI). NDVI and NDBI have theoretical values ranging between -1 and +1 with those > 0 representing vegetated and built-up areas, respectively [25,26]. Both indices are potentially useful for predicting mosquito vector occurrence [26,27]. The remotely sensed data were obtained from the USGS Earth Explorer data portal [24] and included Landsat 8 Operational Land Imager surface reflectance imagery acquired over the study area on 30 July 2017, the most cloud free image acquisition date close to the time of field data collection. Five major land cover types (urban/built-up land, forest land, grassland/agricultural land, barren land, water) were mapped using the random forest classifier at a spatial resolution of 30 m. To define sampling strata, NDVI values were derived for forest land pixels and classified into low (0.472 to 0.859), medium (0.860 to 0.869), and high (0.870 to 1.000) NDVI value categories as described previously [13]. NDBI values were derived for urban/built-up land pixels and classified using the Jenks natural breaks method into low (-1 to -0.467), medium (-0.468 to -0.098), and high (-0.099 to 1.000) NDBI value categories. Several additional NDBI-related variables were calculated to investigate associations between the occurrence of mosquito species and built-up areas. These were the distance from each sampling site to the nearest low, medium, and high NDBI pixel; the maximum NDBI value in the 3 x 3-pixel neighborhood around each site, and the mean NDBI value within 50, 100, 500, 1000, and 2000 m radius around each site. All geospatial analyses were conducted in ArcGIS Pro (ESRI, Redlands, California), ENVI 5.5 (Harris Geospatial, Boulder, California), and R.
Site selection and randomization
Buffer distances were drawn at the forest edge (0 m), and 500, 1000, and 2000 m into the forest. Initially, nine potential sampling sites were randomly generated within each distance and NDVI category (= 108 sites). A further 293 potential sampling sites were generated using the same methods, as required. In addition, 179 potential sampling sites were randomly generated along the southwestern edge of the Ducke reserve bordering the city to intensify sampling in proximity to humans. The latter sites were not stratified by NDVI category, although NDVI values were extracted from all sampling sites for analysis. On a given day, forest edge and forest interior sites were either sampled simultaneously, or sampling alternated daily between the edge and interior, depending on logistics of accessing sites. Interior sites were selected from those available within approximately 5 km of an access point (Fig 2). In total, 244 of the 587 potential sites were sampled over the course of the study, some of which were sampled twice. Of the unique sites, 196 were stratified by distance from forest edge and NDVI category while the remainder were randomly generated along the southwestern edge of the reserve.
The map shows the Ducke reserve and contiguous forest categorized by Normalized Difference Vegetation Index (NDVI) at a 30 m spatial resolution, while the surrounding areas are categorized by Normalized Difference Built-up Index (NDBI) at the same resolution. Both layers were derived from Landsat 8 Operational Land Imager surface reflectance imagery obtained from the USGS Earth Explorer data portal [24]. Sites sampled (black dots) are shown for each distance (continuous lines). The main access points are in the north and south of the reserve.
Mosquito collections
Mosquitoes were usually sampled twice per week across two rainy seasons from January to June 2019, and November 2019 to April 2020, using the following mechanized and human-operated methods (Table 1): BG-Sentinel 2 traps (BioQuip, Rancho Dominguez, California) baited with dry ice and BG-Lure [15], 20 cm diameter hand-nets, and Prokopack aspirators (John W. Hock Company, Gainesville, Florida). BG-Sentinel traps and aspirators were used at sampling sites for the entire study while hand-net collections were initiated one month after the start of the study. Each sampling site was located using a Garmin Rino 755t (Garmin Ltd., Olathe, Kansas) handheld GPS. BG-Sentinel traps were prepared as described previously [13,18] and left to run until the next day. If traps failed (usually due to battery failure), data were excluded from the BG-Sentinel trap dataset (S1 Dataset). Hand-nets were used to manually sample all approaching mosquitoes for approximately 10 minutes while standing within 5 m of traps. Aspirators were used to sample mosquitoes from all vegetation up to chest height within a 5 m radius of traps. We generally sampled with hand-nets on the day traps were placed and with aspirators on the day traps were collected. BG-Sentinel traps were used regardless of weather. Heavy rainfall occasionally limited the use of hand-nets and aspirators, although we were usually able to utilize these on one of the two days associated with each trap collection. Mosquitoes were stored on dry ice at the time of collection and were transferred to a -80°C freezer at the Fundação de Medicina Tropical Doutor Heitor Vieira Dourado (FMT-HVD) until identified.
In situ environmental variables
In situ environmental variables were recorded at each site to investigate associations with distance from edge and the occurrence (presence/absence) of mosquito species. Temperature (°C) and relative humidity (%) were recorded at 30-min intervals for the duration of each BG-Sentinel trap collection using Hygrochron iButton data loggers (Maxim Integrated, San Jose, California) placed in a nylon mesh bag (BioQuip, Rancho Dominguez, California) and attached to the outside of each trap. The minimum, maximum, mean, daytime mean (06:00–18:00), and range of both temperature and relative humidity were calculated for BG-Sentinel trap collections. Values for hand-nets and aspirators were extracted from the datapoint nearest to the time of the collection. Elevation (m) values for each sampling site were calculated as previously described [13]. The % canopy cover was recorded using a spherical densiometer (Forest Densiometers, Rapid City, South Dakota) based on readings made directly above the trap and 5 m from the trap in N, SE, and SW directions. The number of palms with open fronds and the number of mature trees within a 5 m radius of the trap was recorded. Mature trees were defined as those other than palms with a diameter at the base of the trunk of approximately 30 cm or more. The % ground vegetation cover (ground cover) below chest height was estimated in 1 m2 transects at distances 5 m from the trap in N, S, E, and W directions, and expressed as a value: 1 = 0–25%, 2 = 25–50%, 3 = 50–75%, and 4 = 75–100% total cover. Average ground cover was then calculated as the sum of these values, divided by four. Trap substrate was recorded as the estimated % leaf litter, % soil, % grass, % palm frond, % sand, and % other vegetation, present directly beneath the trap, with the total of each type combined to equal 100%. A measure of the slope (°) representative of the general area where the trap was placed was made using the Rotating Sphere Inclinometer v1.8 smartphone app (Calmatics, Gothenburg, Sweden) and the aspect of the slope (°) was recorded using a Garmin Rino 755t handheld GPS.
Mosquito identification
Mosquitoes were placed on a chill table (BioQuip, Rancho Dominguez, California, USA) and identified to species using a stereomicroscope and taxonomic keys as previously described [18]. Genus and species names follow Wilkerson et al. [29]. Samples were stored at -80°C for future arbovirus screening.
Statistical analysis
Contingency tables and two-tailed Fisher’s exact tests were used to compare the genus-level occurrence of Haemagogus and Sabethes mosquitoes across sampling methods between forest edge and interior sites. For this comparison, a single sample bout (i.e., where all sampling methods were employed within ≤14 days) was randomly downselected from each site that was sampled twice, and a very small number of sites from which both genera were collected were excluded. To analyze patterns of collection across the two rainy seasons, data were grouped into three-month sampling blocks using the BG-Sentinel trap dataset (1 = January–March 2019; 2 = April–June 2019; 3 = November 2019 –January 2020; 4 = February–April 2020). Data were grouped this way to generate replicates for analysis while maintaining an acceptably balanced sample size among groups. Mann-Whitney U tests followed by Wilcoxon Each Pair tests were used to compare genus-level abundance across sampling blocks. Two-tailed Fisher’s exact tests were used to compare species-level occurrence across sampling blocks. For the latter, species with excess zero counts were excluded. For both comparisons, a Bonferroni correction was used to adjust alpha for multiple tests.
Analyses of community similarity, richness, and diversity were only performed using samples identified to species level. To compare mosquito communities across distances into the forest, a principal components analysis of relative frequencies of each species was conducted, followed by hierarchical clustering of communities at each distance, as in our previous studies [17,18]. The Morisita overlap index was also used to compare community composition at each distance using the PAST software package [30]. To examine whether sampling was adequate to capture total species richness, iNEXT [31] (R version 4.2.1) was used to generate species accumulation and rarefaction curves by distance for each sampling method as well as for all methods combined (223 sites where all three methods were used on the same sampling occasion). The Shannon-Wiener diversity index was calculated using data grouped in three-month blocks as described above and ANOVA, followed by a Tukey HSD post-hoc test, was used to detect differences in diversity between distances for each sampling method as well as for all methods combined.
The BG-Sentinel trap dataset was used to measure associations of environmental variables with each other and with distance into forest, as this was the largest of the assembled datasets. A single sample was randomly downselected from each site sampled twice, resulting in 235 unique sites for analysis. A Spearman’s rank correlation was then used to test associations between environmental variables, of which six were selected for further analysis based on the generated matrix and our previous work [13]. These were: elevation, maximum temperature, NDVI value, average ground cover, number of palms, and number of mature trees. To compare these variables among the four distances sampled, ANOVA was used when data were normally distributed followed by a Tukey HSD post-hoc test to detect pairwise differences, while the Kruskal-Wallis test was used for non-normal data followed by a Steel-Dwass all-pairs comparison to detect pairwise differences.
Contingency tables and two-tailed Fisher’s exact tests were used to test species occurrence across sites sampled twice for sufficiently common taxa (Aedes, Ae. albopictus, Culex, Limatus, Li. durhamii, Ps. amazonica, Sabethes, Trichoprosopon, and Wyeomyia), to determine whether these sites could be treated as independent samples or needed to be nested. All sample data for each trapping method were then used in nominal logistic regressions to test associations between distance from forest edge (as a categorical variable) and six selected, continuous environmental variables with the occurrence of key taxa. Variables were removed sequentially from the model until all remaining variables contributed significantly or only one variable was left. Analyses of the BG-Sentinel trap dataset were also repeated using only single samples per site, which were randomly downselected as described above.
Since the study aimed to better understand the potential exchange of pathogens at ecosystem boundaries, a Mann-Whitney U test was used to test the association of NDBI values in a 100 m radius around the sampling site with the occurrence of key vector taxa. This radius was chosen for its relevance to average ranges of mosquito movement, but it was significantly correlated with all other NDBI variables. A single sample bout was downselected from sites sampled twice and detection was combined for all three sampling methods, resulting in a total of 100 sites for analysis.
Statistical analyses were performed with JMP Pro 15 [32] unless otherwise stated.
Results
Relative mosquito abundance and general patterns of collection
A total of 9,467 mosquitoes were sampled across 244 unique sites, of which several were sampled twice (detailed below), using BG-Sentinel traps, hand-nets, and Prokopack aspirators (S1 Dataset). The relative abundance and composition of mosquitoes varied by sampling method and distance from the forest edge (Fig 3). All specimens were identified to the rank of genus, and often subgenus, but poor sample condition sometimes prevented species-level identification. Specimens collected with hand-nets were in excellent condition and 97.8% were identified to species level, while those collected with aspirators (80.6% identified to species level) and BG-Sentinel traps (65.0%) were more likely to be damaged. Overall, collections yielded 16 genera and 57 identified species. BG-Sentinel traps (N = 294 sites sampled in total, including 59 sites sampled twice) yielded the most mosquitoes (7,154), comprising 15 genera and 38 identified species, of which 94.4% were female; 1,606 mosquitoes were sampled with hand-nets (N = 256 sites, including 24 sites sampled twice, 12 genera, 40 species, 89.7% female), and 707 mosquitoes were sampled with Prokopack aspirators (N = 287 sites, including 51 sites sampled twice, 13 genera, 35 species, 65.1% female).
Stacks ordered by genus abundance and then alphabetically by subgenus and species. Number of mosquitoes (Mosquitoes n =) and number of collections (Collections n =) made at each distance shown above bar. Number of individuals per taxon included in parentheses next to corresponding name; sp. = single species, spp. = potentially multiple species. Abbreviated names are given in full in the dataset (S1 Dataset). Dendrograms below each graph show hierarchical clustering of principal component (PC)1 and PC2 from a principal components analysis of relative species frequency for each collection method at each distance sampled.
Psorophora was the most abundant genus sampled with BG-Sentinel traps and hand-nets, while Aedes was the most abundant genus sampled with aspirators (Fig 3). Psorophora amazonica and Ae. albopictus were the dominant species within respective genera and both were sampled in high relative abundance using each method. Both Psorophora generally and Ps. amazonica specifically were present in higher relative abundance in the forest interior than at the edge, while Ae. albopictus was predominantly found at the forest edge, occasionally at 500 m, and rarely at 1000 m into the forest. Culex species were present in high relative abundance in BG-Sentinel traps and aspirators but were unlikely to be biting humans in high numbers during the daytime since they were seldom collected in hand-nets. Sabethes species were more common in BG-Sentinel traps than hand-nets while the opposite was true for Haemagogus generally and Hg. janthinomys specifically. Neither Sabethes nor Haemagogus were common in aspirator collections. Contingency table analyses of species presence at each site revealed that Sabethes mosquitoes (all species combined) were more common than Haemagogus mosquitoes (all species combined) at the edge and less common in interior sites when either net collections (Fisher’s exact test, DF = 2, χ2 = 15.5, P = 0.0004) or BG-Sentinel trap collections (DF = 2, χ2 = 13.0, P = 0.002) were analyzed; too few specimens were collected using aspirators to permit analysis.
Differences in genus-level abundance across three-month sampling blocks were only significant after a Bonferroni correction for multiple testing for Culex (Mann-Whitney U test, DF = 3, χ2 = 13.8, P = 0.003), which was most abundant in blocks 1 and 4, and Trichoprosopon (DF = 3, χ2 = 21.9, P < 0.0001), which was most abundant in blocks 1 and 2. Comparisons of species occurrence showed significant differences for Li. durhamii (DF = 3, χ2 = 14.7, P = 0.002), Ps. amazonica (DF = 3, χ2 = 15.6, P = 0.001), and Tr. digitatum (DF = 3, χ2 = 24.1, P < 0.0001), all of which were more common in month blocks 1 and 2 than in blocks 3 and 4, corresponding with higher occurrence in the first rainy season. There was no difference in the occurrence of Ae. albopictus among sampling blocks (P > 0.05).
Analyses of community similarity, richness, and diversity
In a principal components analysis of the relative frequency of each species at each distance, the first two principal components (PC1 and PC2) explained 89.2% of the variation in the BG-Sentinel trap data, 92.6% of variation in hand-net data, and 80.9% of variation in aspirator data. For BG-Sentinel trap data, PC1 reflected the relative abundance of a group of 17, predominantly rare species, though this group did include Li. durhamii and Ae. albopictus, while PC 2 reflected the relative abundance of six additional, rare species. Generally, the loadings were similar for the aspirator data, albeit PC1 also reflected the relative abundance of Ps. amazonica. However, loadings were strikingly different for the hand-net data, where PC1 reflected the relative abundance of Hg. janthinomys, while PC2 reflected the abundance of four rare species that did not, with one exception (Ps. cingulata in BG-Sentinel trap data), load heavily onto PC1 or PC2 in the other two datasets. Hierarchical cluster analysis of PC1 and PC2 (Fig 3) revealed that, for all sampling methods, communities at 1000 m and 2000 m were quite similar but differed substantially from communities at 500 m, which in turn differed from communities at the edge.
The Morisita overlap index (S1 Table) derived from hand-net and aspirator collections showed that forest edge communities differed from those sampled at 500, 1000, and 2000 m (Morisita index < 0.32 for all comparisons), while interior distances were quite similar when compared with each other (> 0.80 for all comparisons). When this index was calculated using BG-Sentinel trap data, communities at 500 m and 1000 m were more distinct from edge communities than those at 2000 m. However, this disparity was due to large numbers of Cx. coronator collected in traps at 2000 m over two days in January 2019 (99.8% of the total Cx. coronator collected at this distance). If removed, the results followed a similar trend to both hand-nets and aspirators.
Species accumulation and rarefaction curves showed that species richness was generally higher at 0 m and 500 m than at 1000 m and 2000 m across the three sampling methods and for all methods combined (Fig 4). Results were more difficult to interpret for Prokopack aspirators, which were relatively ineffective inside the forest where fewer sites were sampled at each distance compared to forest edges. Overall, species were well represented at 0, 1000, and 2000 m, although some rarer species remained to be sampled. The steep increase in accumulation curves at 500 m shows that additional sampling was needed to capture the total richness at this distance.
Panels show species richness by distance for each sampling method, and combined data for 223 sites where all three methods were used on the same sampling occasion. Shaded areas surrounding rarefaction and extrapolation lines represent 95% confidence intervals. Curves generated using taxa identified to species level only.
Species diversity, as measured by the Shannon-Wiener diversity index, differed significantly across distance from forest edge when sampling with hand-nets (ANOVA, DF = 3, F = 8.06, P = 0.003) and for all methods combined (ANOVA, DF = 3, F = 6.83, P = 0.006). For hand nets, these differences were significant between 0 m and 1000 m, 0 m and 2000 m, and 500 m and 2000 m (Tukey HSD, P < 0.05 for all comparisons), and for all methods combined, differences were significant between 0 m and 1000 m, 0 m and 2000 m, and 500 m and 1000 m (Tukey HSD, P < 0.05 for all comparisons). Diversity was generally higher at 0 m and 500 m than at 1000 m and 2000 m (S2 Table). Differences in diversity were not significant when sampling with BG-Sentinel traps (ANOVA, DF = 3, F = 1.94, P = 0.18) or aspirators (ANOVA, DF = 3, F = 2.30, P = 0.13), although both showed similar trends for higher diversity towards the forest edge.
Differences in environmental variables by distance into forest
A Spearman’s rank correlation showed that measures of NDVI and NDBI were generally inversely correlated, measures of microclimate were highly correlated, and most of the environmental variables measured, except for elevation, were correlated to some degree (S3 Table). Of the six environmental variables selected for further analysis (elevation, maximum temperature, NDVI value, average ground cover, number of palms, and number of mature trees), elevation increased slightly (15 m) but significantly (ANOVA, DF = 3, F = 5.85, P = 0.0007) between 0 m and 500 m into the forest; there was no significant difference in elevation between sites at 500, 1000, and 2000 m (S4 Table). NDVI followed a similar trend, increasing significantly (Kruskal-Wallis, DF = 3, χ2 = 60.7, P < 0.0001) between 0 m and 500 m and remaining steady thereafter. Forest edge sites were hotter and less humid than each of the interior distance categories (Fig 5). Maximum temperature decreased significantly between 0 m and 500 m (Kruskal-Wallis test, DF = 3, χ2 = 29.3, P < 0.0001) and remained stable at 500, 1000, and 2000 m. Average ground cover decreased continuously from 0 m to 1000 m and subsequently leveled off (Kruskal-Wallis, DF = 3, χ2 = 30.3, P < 0.0001). The number of palms and mature trees increased abruptly between 0 m and 500 m and leveled off across the remaining distance categories (Kruskal-Wallis, DF = 3, χ2 = 96.1 for palms and 59.3 for mature trees, P < 0.0001 for both comparisons).
Left panel shows daily cycles of temperature (°C) and relative humidity (%) recorded at 30-min intervals for each distance from forest edge. Data points derived from all BG-Sentinel trap collections where temperature and relative humidity were recorded, including those where traps failed (S1 Dataset). Right panel shows mean minimum relative humidity (%) and mean maximum temperature (°C) ± 1 standard error (S.E.) at each distance sampled, constructed using data from the final BG-Sentinel trap dataset where traps were still working upon collection.
Associations between select variables and the occurrence of key taxa
Among the common taxa sampled with BG-Sentinel traps (Aedes, Ae. albopictus, Culex, Limatus, Li. durhamii, Ps. amazonica, Sabethes, Trichoprosopon, and Wyeomyia), only Limatus showed a significant association (P = 0.008) between the first and second sample at the same site after a Bonferroni correction. This comparison remained significant when limited to just the 0 m distance. No associations were significant for hand-net collections, and aspirators generally collected too few samples for analysis. The average time between the first and second sample was 84 days (min. = 40 d, max. = 143 d).
Nominal logistic regressions used to test associations between the six selected environmental variables and the occurrence of key taxa showed that three taxa were associated with forest edges (Table 2). Of these, Ae. albopictus showed strong negative associations with distance from forest edge and positive associations with elevation across all sampling methods, in addition to negative associations with NDVI value and the number of mature trees across one of the three sampling methods. Aedes scapularis showed negative associations with distance from forest edge, temperature, and the number of mature trees, and positive associations with elevation and NDVI value across at least one of the three sampling methods. Limatus durhamii showed strong negative associations with distance from forest edge and positive associations with NDVI value and average ground cover, although the latter was only marginally significant. Of the forest specialists, Ps. amazonica showed positive associations with distance from forest edge, elevation, NDVI value, and the number of mature trees, and a negative association with temperature. However, each of these associations was only significant for one of the three sampling methods, despite Ps. amazonica being the most frequently encountered species. Haemagogus generally and Hg. janthinomys specifically showed positive associations with distance from forest edge. In contrast, there was no association between distance from forest edge and the occurrence of either subgenus of Sabethes mosquitoes, although Sabethes (Sabethes) spp. showed positive associations with temperature and average ground cover.
Blue and red shaded cells indicate significant positive or negative associations, respectively. Gray shaded cells represent factors that were not significantly associated with the designated taxon.
Associations between surrounding levels of NDBI and the occurrence of key taxa
Sites in which Ae. aegypti (Mann-Whitney U test, N = 3 positive sites, DF = 1, χ2 = 6.2, P = 0.01) and Ae. albopictus (N = 65 positive sites, DF = 1, χ2 = 19.7, P < 0.0001) were detected had significantly higher mean NDBI values in a 100 m radius around the site than sites where these species were not detected. Sabethes mosquitoes showed an opposite trend, with positive sites (N = 22) having lower levels of NDBI within 100 m, albeit the difference was only marginally significant (χ2 = 3.7; P = 0.06). However, sites in which Haemagogus (N = 6) or Psorophora mosquitoes (N = 50) were detected did not differ in mean NDBI value of the surrounding area when compared to sites where they were not detected (P > 0.48 for both comparisons). S5 Table shows the mean distance to low, medium, and high NDBI pixels for sites in which select taxa were detected.
Discussion
This study expands upon our earlier investigation of ground-level changes in mosquito community composition at the edges of urban forest fragments in Manaus, where we aimed to identify potential routes of spillover and spillback of mosquito-borne viruses [13]. We previously focused on relatively small (≤6.6 km2) forest fragments embedded within the urban matrix, bordering densely populated areas, where humans, mosquitoes, and wildlife could readily interact. In contrast, the Ducke reserve covers 100 km2 of primary rainforest [19] and while its southwestern edge forms a striking interface with the city, the remaining edges border rural areas which are more or less contiguous with the surrounding Amazon rainforest.
We used three complimentary sampling methods to maximize representation of mosquito fauna in the study area. Species abundance was highest among BG-Sentinel trap collections (75.6% of all mosquitoes sampled), which was expected and agrees with other work in forests utilizing multiple sampling methods [16,33]. However, these data were somewhat inflated by relatively rare sampling events of large numbers of Psorophora and Culex mosquitoes (S1 Dataset). The overall richness of identified species was similar across the three sampling methods. Species richness was similarly high at the forest edge among BG-Sentinel trap, hand-net, and aspirator collections, and was higher towards the edge (0 m and 500 m) for all methods combined. Species richness was generally lower at interior sites (1000 m and 2000 m), particularly among hand-net collections, where fewer anthropophilic species were present and Ps. amazonica dominated. Aspirators were relatively ineffective when sampling mosquitoes inside the forest, likely due to the wealth of available resting sites, particularly above the ground.
When considering key vectors, each sampling method showed Ae. albopictus to be present in high relative abundance at forest edges. This species was also present 500 m and 1000 m into the forest, but only in very low numbers, agreeing with studies conducted at the same reserve [34] and in the Tijuca Forest, Rio de Janeiro [35]. We rarely encountered Ae. aegypti, which was only sampled at three sites along the urban edge. Haemagogus and Sabethes mosquitoes formed less than 2% of the overall catch which was expected since these genera comprise mostly canopy-dwelling species [18]. Whereas Haemagogus mosquitoes readily approach humans [36], Sabethes approach more slowly and are easily disturbed by movement [37], a behavior we have repeatedly observed in the field. Consequently, short duration hand-net collections, which were effective at detecting Hg. janthinomys, seemed less effective for sampling Sabethes. In contrast, longer duration collections with dry-ice-baited BG-Sentinel traps seemed better for this purpose.
The vector status of Ps. amazonica is currently unknown. However, it is a highly seasonal, highly anthropophilic species, which is mainly present during the latter months of the rainy season and has been well represented in each of our previous studies [13,17,18]. In the present study, Ps. amazonica formed 92.2% of the identified Psorophora, and was particularly abundant in hand-net collections, reinforcing its status as an aggressive human biter. Crucially, its substantial presence at each distance and height [17] sampled provides a clear potential pathway for pathogen exchange between the forest edge and canopy. Vector competence studies involving this species are clearly needed.
We found differences in the abundance and occurrence of certain taxa across three-month sampling blocks and between rainy seasons. Trichoprosopon was the only genus with higher abundance in month blocks 1 and 2, corresponding to the first rainy season, compared to month blocks 3 and 4. The occurrence of Li. durhamii, Ps. amazonica, and Tr. digitatum was also significantly higher during month blocks 1 and 2. We previously observed peaks in the abundance and occurrence of Limatus, Psorophora, and Trichoprosopon during April and May when sampling smaller forest fragments in Manaus [13]. Since our sampling was truncated midway through the second rainy season by the COVID-19 pandemic, we likely missed periods of peak activity for some of these taxa.
Overall, forest edges differed in richness and diversity of mosquito communities compared to the forest interior, with communities at 1000 m and 2000 m being most alike and least diverse. Some groups of organisms are thought to be diverse and species rich at forest edges where interior and exterior species coexist along with edge specialists [11,38,39]. In this regard, mosquitoes have not been extensively studied and results of field investigations are inconclusive. Steiger at al. [11,12] reported contrasting findings when analyzing species richness across grassland, edge, and interior habitats of an Australian tropical forest, but noted significant changes in community composition between the edge and interior in both studies. We found no difference in species richness and diversity when sampling up to 500 m from the forest edge in our previous study [13], but like Steiger et al. [11,12], we detected substantial changes in community composition. Young et al. [40] conducted a more intricate study of mosquito communities at forest edges in Malaysian Borneo, finding little difference in richness, diversity, and community composition when sampling interior distances up to 500 m from the edge. We sampled deeper into the forest and at a coarser resolution in the present study, where we generally found a peak in richness and diversity towards the edge, which declined beyond 500 m. Species accumulation curves showed that rarer taxa remained to be sampled, particularly at 500 m, although these are less likely to serve as bridge vectors given their relative scarcity. Similar results from other studies highlight the difficulty of sampling mosquito species to saturation at ground level in forested environments [40,41], especially given biases associated with different sampling methods [16,33,42].
Analyses of the six selected environmental variables showed that changes mainly occurred between the forest edge and 500 m. Thereafter, conditions remained relatively stable, apart from average ground cover which gradually declined up to 1000 m. These findings are in broad agreement with the permeability of forest edges to external biotic and abiotic factors elsewhere near Manaus [39]. We found that NDVI value was lowest at forest edges and increased significantly by 500 m, agreeing with our earlier study [13] and others who have shown similar changes within 150 m of urban-forest edges [43]. Field observations have shown that desiccating effects can penetrate up to 200 m into forests [39], altering microclimate by creating hotter and drier conditions [13]. Satellite observations have shown that these effects can penetrate beyond 1 km in moderately and highly fragmented forest [44]. Tree mortality is often elevated at forest edges, disproportionally affecting larger trees [45] and adult palms [46], which may lead to increased light penetration and successional plant growth [47]. In accordance, we detected an increase in the number of mature trees and palms, and a decrease in average ground cover moving from the forest edge to interior. We also detected a slight (15 m) increase in elevation between the edge and 500 m. While changes in topography may influence abiotic conditions and vegetation characteristics [48], the effects are likely to be minimal given the small difference we encountered. Many of the above findings agree with the results of our previous study in urban forest fragments in Manaus [13].
Analyzing associations between distance from forest edge and the six specified environmental variables with species occurrence highlighted several traits associated with key taxa. Aedes albopictus was negatively associated with distance from forest edge across all sampling methods, as well as with NDVI value and the number of mature trees when sampled with nets and traps, respectively, agreeing with its preference for edge habitats among forested areas [13,34]. Although generally considered a peri-urban species which occasionally breeds in water-gathering tree holes [49], dense primary vegetation may not represent ideal habitat for Ae. albopictus. Other species associated with forest edges included Ae. scapularis and Li. durhamii, which were mainly found within 500 m of the border. The former showed similar trends to Ae. albopictus but was negatively associated with temperature and positively associated with NDVI value across one of the three sampling methods, suggesting it may be better adapted to areas of more dense vegetation. Indeed, Ae. scapularis is commonly found in forests [50,51] but is also well adapted to human environments [52] and has potential to transmit multiple viruses of public health importance including YFV [53,54] and MAYV [55]. Limatus durhamii showed strong negative associations with distance from forest edge and positive associations with NDVI value and average ground cover, although the association with the latter was only weak. This species breeds in both natural and artificial containers [56] and adult females are attracted to humans [51], but there is currently only limited evidence of a role for Limatus durhamii in arbovirus transmission [57].
Psorophora amazonica showed positive associations with several variables suggestive of forest-dwelling species (distance from forest edge, NDVI value, and the number of mature trees), but these associations were not consistent across sampling methods. This was true for most variables, other than distance from forest edge, associated with the key taxa listed in Table 2. We detected a negative association between Ps. amazonica and temperature when sampling with hand-nets. We previously found a negative association with temperature when sampling Ps. albigenu, but not Ps. amazonica. It seems that Psorophora (Janthinosoma) species are generally more abundant during cooler conditions (< 26°C) than their forest-dwelling counterparts, although they tolerate a wide range of temperatures [17]. Of the variables analyzed, Haemagogus generally and Hg. janthinomys specifically showed positive associations with distance from forest edge. These taxa were sampled in relatively low numbers at ground level, which may be why we failed to detect associations with temperature and humidity that we recorded previously when sampling vertically [17,18]. Since Hg. janthinomys breeds mainly in tree holes in primary tropical rainforest [36,58], the dramatic environmental shifts occurring between 0 m and 500 m, where large tree mortality is elevated [45], may explain its relative scarcity compared to Sabethes species mosquitoes. The latter, which breed among a range of phytotelmata including bamboo internodes [59,60], may be better suited to forest edges. In support of this, we found no associations between the occurrence of Sabethes taxa and distance from forest edge at subgenus level, but we did find positive associations between Sabethes (Sabethes) spp., temperature, and average ground cover, which are higher at forest edges.
Zoonotic exchange of DENV, ZIKV, CHIKV, YFV, and MAYV requires spatiotemporal overlap of humans, susceptible monkeys (or other wildlife hosts), and competent mosquito vectors at or near forest edges. The proximity of human habitations to forest edges will influence the presence of non-native species in respective habitats (i.e., forest mosquitoes in human habitations and urban mosquitoes at forest edges). We used NDBI as a proxy for urbanization and found that sites positive for Ae. aegypti and Ae. albopictus had significantly higher mean NDBI values within a 100 m radius than sites that were not positive for these species. The mean distance of positive sites to the nearest medium NDBI pixel was similar for both species (S5 Table), while the mean distance to the nearest high NDBI pixel was greater for Ae. albopictus, in keeping with their respective preferences for urban and rural areas [61]. The greater mean distance of Sabethes, Haemagogus, and Psorophora positive sites to medium and high NDBI pixels suggests that these taxa are more likely to be encountered at forest edges bordering peri-urban or rural areas.
We sampled a forest reserve managed by the National Institute of Amazonian Research (INPA) where access is officially restricted. However, anecdotally, people do occasionally enter the reserve including for recreation, gathering fruit, and hunting. Illegal deforestation for the establishment of “invasion” communities, including the former Monte Horebe community on the western edge of the reserve [62], not only brings people into extremely close contact with recently cleared rainforest, but likely removes physical and social barriers associated with entering forests that exist in pre-planned urban areas. Behavioral studies are needed to better understand relationships between residents and forests bordering Manaus, and to better understand the risk of the spillover and spillback in these areas.
Conclusion
Our findings suggest that major changes in environmental variables occur within 500 m of the forest edge at the Ducke reserve, where there is high risk for contact with both urban and sylvatic vectors, and that these are associated with substantial shifts in mosquito diversity and community composition. By 1000 m, conditions stabilize, species diversity decreases, and forest mosquitoes predominate. Environmental variables associated with the occurrence of key taxa may be leveraged to characterize suitable habitat and refine risk models for pathogen spillover and spillback.
Supporting information
S1 Dataset. Datasets for each sampling method and environmental variables.
https://doi.org/10.1371/journal.pntd.0011296.s001
(XLSX)
S3 Table. Spearman’s rank analysis of environmental variables.
https://doi.org/10.1371/journal.pntd.0011296.s004
(DOCX)
S4 Table. Associations between select environmental variables and distance.
https://doi.org/10.1371/journal.pntd.0011296.s005
(DOCX)
S5 Table. Associations between NDBI and key taxa.
https://doi.org/10.1371/journal.pntd.0011296.s006
(DOCX)
Acknowledgments
We dedicate this manuscript to the memory of Nelson Ferreira Fé, a talented entomologist, taxonomist, and our friend, who contributed immensely to vector-borne disease research in the Brazilian Amazon. We wish to acknowledge Ramon Linhares and Layne Borges (FMT-HVD) for assistance with sample processing; Raimundo Coelho and Aleksey Lima (FMT-HVD) for driving and logistics, and the Instituto Nacional de Pesquisas da Amazônia (INPA) for permission to use the Adolpho Ducke Reserve to collect samples for this study.
References
- 1. Despommier D, Ellis BR, Wilcox BA. The role of ecotones in emerging infectious diseases. EcoHealth. 2006;3(4):281–9.
- 2. Borremans B, Faust C, Manlove KR, Sokolow SH, Lloyd-Smith JO. Cross-species pathogen spillover across ecosystem boundaries: mechanisms and theory. Philos Trans R Soc Lond B Biol Sci. 2019;374(1782):20180344. pmid:31401953
- 3. Weaver SC, Charlier C, Vasilakis N, Lecuit M. Zika, Chikungunya, and Other Emerging Vector-Borne Viral Diseases. Annu Rev Med. 2018;69:395–408. pmid:28846489
- 4. Wilder-Smith A, Gubler DJ, Weaver SC, Monath TP, Heymann DL, Scott TW. Epidemic arboviral diseases: priorities for research and public health. Lancet Infect Dis. 2017;17(3):e101–e6. pmid:28011234
- 5. Weaver SC, Reisen WK. Present and future arboviral threats. Antiviral Res. 2010;85(2):328–45. pmid:19857523
- 6. Bryant JE, Holmes EC, Barrett AD. Out of Africa: a molecular perspective on the introduction of yellow fever virus into the Americas. PLoS Pathog. 2007;3(5):e75. pmid:17511518
- 7. Althouse BM, Vasilakis N, Sall AA, Diallo M, Weaver SC, Hanley KA. Potential for Zika virus to establish a sylvatic transmission cycle in the Americas. PLoS Negl Trop Dis. 2016;10(12):e0005055. pmid:27977671
- 8. Lourenço-de-Oliveira R, Failloux AB. High risk for chikungunya virus to initiate an enzootic sylvatic cycle in the tropical Americas. PLoS Negl Trop Dis. 2017;11(6):e0005698. pmid:28662031
- 9. Hanley KA, Monath TP, Weaver SC, Rossi SL, Richman RL, Vasilakis N. Fever versus fever: The role of host and vector susceptibility and interspecific competition in shaping the current and future distributions of the sylvatic cycles of dengue virus and yellow fever virus. Infect Genet Evol. 2013;19:292–311. pmid:23523817
- 10. Mbaika S, Lutomiah J, Chepkorir E, Mulwa F, Khayeka-Wandabwa C, Tigoi C, et al. Vector competence of Aedes aegypti in transmitting Chikungunya virus: effects and implications of extrinsic incubation temperature on dissemination and infection rates. Virol J. 2016;13(1):114. pmid:27357190
- 11. Steiger DM, Ritchie SA, Laurance SGW. Mosquito communities and disease risk influenced by land use change and seasonality in the Australian tropics. Parasit Vectors. 2016;9(1):387–. pmid:27388293
- 12. Steiger DM, Johnson P, Hilbert DW, Ritchie S, Jones D, Laurance SG. Effects of landscape disturbance on mosquito community composition in tropical Australia. J Vector Ecol. 2012;37(1):69–76. pmid:22548538
- 13. Hendy A, Hernandez-Acosta E, Chaves BA, Fe NF, Valerio D, Mendonca C, et al. Into the woods: Changes in mosquito community composition and presence of key vectors at increasing distances from the urban edge in urban forest parks in Manaus, Brazil. Acta Trop. 2020;206:105441. pmid:32173316
- 14. Vazquez-Prokopec GM, Galvin WA, Kelly R, Kitron U. A new, cost-effective, battery-powered aspirator for adult mosquito collections. J Med Entomol. 2009;46(6):1256–9. pmid:19960668
- 15.
Biogents AG. The BG-Sentinel: Biogents’ mosquito trap for researchers 2020 [cited 2020]. Available from: https://www.bg-sentinel.com/.
- 16. Obenauer PJ, Kaufman PE, Kline DL, Allan SA. Detection of and monitoring for Aedes albopictus (Diptera: Culicidae) in suburban and sylvatic habitats in north central Florida using four sampling techniques. Environ Entomol. 2010;39(5):1608–16. pmid:22546459
- 17. Hendy A, Valério D, Fé NF, Hernandez-Acosta E, Mendonça C, Andrade E, et al. Microclimate and the vertical stratification of potential bridge vectors of mosquito-borne viruses captured by nets and ovitraps in a central Amazonian forest bordering Manaus, Brazil. Sci Rep. 2021;11(1):21129. pmid:34702887
- 18. Hendy A, Hernandez-Acosta E, Valério D, Mendonça C, Costa ER, Júnior JTA, et al. The vertical stratification of potential bridge vectors of mosquito-borne viruses in a central Amazonian forest bordering Manaus, Brazil. Sci Rep. 2020;10(1):18254. pmid:33106507
- 19. Oliveira MLd Baccaro FB, Braga-Neto R Magnusson WE. Reserva Ducke: A bioversidade Amazônica através de uma grade. Manaus: Instituto Nacional de Pesquisas da Amazônia; 2011. 166 p.
- 20. Walsh JF, Molyneux DH, Birley MH. Deforestation: effects on vector-borne disease. Parasitology. 1993;106 Suppl:S55–75. pmid:8488073
- 21. Gonçalves Maciel LH, Vieira da Rocha Neto, Ferreira Martins Y, de Azevedo Furtado F, Cunha Teixeira P, Oliveira Dias MY, et al. Prevalence of arboviruses and other infectious causes of skin rash in patients treated at a tertiary health unit in the Brazilian Amazon. PLoS Negl Trop Dis. 2022;16(10):e0010727. pmid:36228027
- 22. Saraiva MdGG, Amorim RDS, Moura MAS Santos ECSd, Sampaio LS, Barbosa MdGV, et al. Historical analysis of the records of sylvan yellow fever in the State of Amazonas, Brazil, from 1996 to 2009. Rev Soc Bras Med Trop. 2013;46:223–6. pmid:23740065
- 23. Mourão MP, Bastos MdS, Figueiredo RM, Gimaque JB, Alves VdC, Saraiva Md, et al. Arboviral diseases in the Western Brazilian Amazon: A perspective and analysis from a tertiary health & research center in Manaus, State of Amazonas. Rev Soc Bras Med Trop. 2015;48:20–6. pmid:26061367
- 24.
United States Geological Survey. Earth explorer 2017 [cited 2017]. Available from: https://earthexplorer.usgs.gov/.
- 25. Liu H, Weng Q. Enhancing temporal resolution of satellite imagery for public health studies: A case study of West Nile Virus outbreak in Los Angeles in 2007. Remote Sens Environ. 2012;117:57–71.
- 26. Estallo EL, Sangermano F, Grech M, Ludueña-Almeida F, Frías-Cespedes M, Ainete M, et al. Modelling the distribution of the vector Aedes aegypti in a central Argentine city. Med Vet Entomol. 2018;32(4):451–61. pmid:30027565
- 27. Richman R, Diallo D, Diallo M, Sall AA, Faye O, Diagne CT, et al. Ecological niche modeling of Aedes mosquito vectors of chikungunya virus in southeastern Senegal. Parasit Vectors. 2018;11(1):255. pmid:29673389
- 28. Amos BA, Cardé RT. Efficiency of CO2-baited CDC miniature light traps under semi-field conditions and characterizing response behaviors of female Aedes aegypti (Diptera: Culicidae). J Vector Ecol. 2020;45(2):180–7. pmid:33207060
- 29.
Wilkerson RC, Linton Y-M, Strickman D. Mosquitoes of the World. Baltimore: Johns Hopkins University Press; 2021. 1332 p.
- 30. Hammer Ø, Harper DAT, Ryan PD. PAST: Paleontological statistics software package for education and data analysis. Palaeontol Electronica. 2001;4(1):9.
- 31. Hsieh TC, Ma KH, Chao A. iNEXT: an R package for rarefaction and extrapolation of species diversity (Hill numbers). Methods Ecol Evol. 2016;7(12):1451–6.
- 32.
JMP Pro 15. Cary, NC, USA: SAS Institute Inc; 2018.
- 33. de Miranda RM, Ferreira-de-Brito A, Silva JdS, Xavier AdS, Freitas Silva SO, Alencar J, et al. Mosquito Fauna and Spatial Distribution in an Atlantic Forest Area in Rio de Janeiro State, Brazil, Reveal a High Risk of Transmission of Yellow Fever and Other Arboviruses. Trop Med Infect Dis. 2022;7(12):410. pmid:36548665
- 34. Pereira Dos Santos T, Roiz D, Abreu FVS, Luz SLB, Santalucia M, Jiolle D, et al. Potential of Aedes albopictus as a bridge vector for enzootic pathogens at the urban-forest interface in Brazil. Emerg Microbes Infect. 2018;7(1):191. pmid:30482898
- 35. Lourenço-de-Oliveira R, Castro MG, Braks MA, Lounibos LP. The invasion of urban forest by dengue vectors in Rio de Janeiro. J Vector Ecol. 2004;29(1):94–100. pmid:15266746
- 36. Arnell JH. Mosquito studies (Diptera, Culicidae) XXXII. A revision of the genus Haemagogus. Contrib Am Entomol Inst. 1973;10(2):1–174.
- 37. Evans AM, Walker GR. Notes on Brazilian mosquitoes: Species observed in the Amazon Valley, and record of Aédes albifasciatus Macq. invading a ship in harbour. Ann Trop Med Parasitol. 1935;29(4):463–7.
- 38. Erdős L, Gallé R, Körmöczi L, Bátori Z. Species composition and diversity of natural forest edges: edge responses and local edge species. Community Ecol. 2013;14(1):48–58.
- 39. Laurance WF, Camargo JLC, Fearnside PM, Lovejoy TE, Williamson GW, Mesquita RCG, et al. An Amazonian rainforest and its fragments as a laboratory of global change. Biol Rev. 2018;93:223–47. pmid:28560765
- 40. Young KI, Buenemann M, Vasilakis N, Perera D, Hanley KA. Shifts in mosquito diversity and abundance along a gradient from oil palm plantations to conterminous forests in Borneo. Ecosphere. 2021;12(4):e03463. pmid:33996190
- 41. Pereira-Silva JW, Ríos-Velásquez CM, Lima GRd, Marialva dos Santos EF, Belchior HCM, Luz SLB, et al. Distribution and diversity of mosquitoes and Oropouche-like virus infection rates in an Amazonian rural settlement. PLoS One. 2021;16(2):e0246932. pmid:33592052
- 42. Mucci LF, Bergo ES, Deus JTd, Reginato SL, Pereira M, Camargo-Neves VLFd. Evaluation of Methods for Collecting Mosquitoes (Culicidae: Diptera) in Canopy and Ground Strata in the Brazilian Savanna. Trop Med Infect Dis. 2022;7(12):446. pmid:36548701
- 43. Bae J, Ryu Y. The magnitude and causes of edge effects on soil organic carbon stocks within and across urban to rural forest patches. Landsc Urban Plan. 2021;215:104223.
- 44. Briant G, Gond V, Laurance SGW. Habitat fragmentation and the desiccation of forest canopies: A case study from eastern Amazonia. Biol Conserv. 2010;143(11):2763–9.
- 45. Laurance WF, Delamônica P, Laurance SG, Vasconcelos HL, Lovejoy TE. Rainforest fragmentation kills big trees. Nature. 2000;404(6780):836–. pmid:10786782
- 46. Baez S, Balslev H. Edge effects on palm diversity in rain forest fragments in western Ecuador. Biodivers Conserv. 2007;16(7):2201–11.
- 47. Laurance WF, Nascimento HE, Laurance SG, Andrade AC, Fearnside PM, Ribeiro JE, et al. Rain forest fragmentation and the proliferation of successional trees. Ecology. 2006;87(2):469–82. pmid:16637371
- 48. Lippok D, Beck SG, Renison D, Hensen I, Apaza AE, Schleuning M. Topography and edge effects are more important than elevation as drivers of vegetation patterns in a neotropical montane forest. J Veg Sci. 2014;25(3):724–33.
- 49. Pereira-dos-Santos T, Roiz D, Lourenço-de-Oliveira R, Paupy C. A Systematic Review: Is Aedes albopictus an Efficient Bridge Vector for Zoonotic Arboviruses? Pathogens. 2020;9(4):266. pmid:32272651
- 50. Galindo P, Trapido H, Carpenter SJ. Observations on diurnal forest mosquitoes in relation to sylvan yellow fever in Panama. Am J Trop Med Hyg. 1950;30(4):533–74. pmid:15425744
- 51. Lira-Vieira AR, Gurgel-Goncalves R, Moreira IM, Yoshizawa MAC, Coutinho ML, Prado PS, et al. Ecological aspects of mosquitoes (Diptera: Culicidae) in the gallery forest of Brasília National Park, Brazil, with an emphasis on potential vectors of yellow fever. Rev Soc Bras Med Trop. 2013;46:566–74. pmid:24142367
- 52. Guimarães AÉ, Motta MdA, Arlé M, Machado RM, Goncalves LD. Bionomics of mosquitoes (Diptera-Culicidae) in areas of the Atlantic Forest of Itaguaí municipality, Rio de Janeiro state, Brazil: I. intra, peri and extradomiciliary frequency. Mem Inst Oswaldo Cruz. 1989;84(suppl 4):243–54.
- 53. Vasconcelos PF, Costa ZG, Travassos Da Rosa ES, Luna E, Rodrigues SG, Barros VL, et al. Epidemic of jungle yellow fever in Brazil, 2000: implications of climatic alterations in disease spread. J Med Virol. 2001;65(3):598–604. pmid:11596099
- 54. Cunha MS, Tubaki RM, de Menezes RMT, Pereira M, Caleiro GS, Coelho E, et al. Possible non-sylvatic transmission of yellow fever between non-human primates in São Paulo city, Brazil, 2017–2018. Sci Rep. 2020;10(1):15751. pmid:32978448
- 55. Aitken THG, Downs WG, Anderson CR, Spence L, Casals J. Mayaro virus isolated from a Trinidadian mosquito, Mansonia venezuelensis. Science. 1960;131(3405):986–.
- 56. Bastos AQ, de Mello CF, dos Santos Silva J, Gil-Santana HR, Silva SOF, Alencar J. Diversity of Mosquitoes (Diptera: Culicidae) in the Bom Retiro Private Natural Heritage Reserve, Rio de Janeiro State, Brazil. J Med Entomol. 2022;59(2):446–53. pmid:35026036
- 57. Barrio-Nuevo KM, Cunha MS, Luchs A, Fernandes A, Rocco IM, Mucci LF, et al. Detection of Zika and dengue viruses in wild-caught mosquitoes collected during field surveillance in an environmental protection area in São Paulo, Brazil. PLoS One. 2020;15(10):e0227239. pmid:33064724
- 58. Bugher JC, Boshell-Manrique J, Roca-García M, Osorno-Mesa E. Epidemiology of jungle yellow fever in eastern Colombia. Am J Epidemiol. 1944;39(1):16–51.
- 59. Talaga S, Dejean A, Carinci R, Gaborit P, Dusfour I, Girod R. Updated Checklist of the Mosquitoes (Diptera: Culicidae) of French Guiana. J Med Entomol. 2015;52(5):770–82. pmid:26336249
- 60. Bastos AQ, Leite PJ, de Mello CF, Maia DA, Machado SL, Gil-Santana HR, et al. Bionomy of mosquitoes in bamboo internodes in an Atlantic Forest remnant of the State of Rio De Janeiro, Brazil. J Am Mosq Control Assoc. 2021;37(4):208–15. pmid:34817616
- 61. Braks MA, Honório NA, Lourenço-de-Oliveira R, Juliano SA, Lounibos LP. Convergent habitat segregation of Aedes aegypti and Aedes albopictus (Diptera: Culicidae) in southeastern Brazil and Florida. J Med Entomol. 2003;40(6):785–94.
- 62. Silvestrim EG, Vieira MRS, Silvestrim FG, Lima Filho AAd, Lopes AB. An insight into the environmental degradation caused by irregular occupations in green areas in the city of Manaus-AM. Research, Society and Development. 2021;10(16):e290101623764.