Portal de Periódicos da CAPES

Effects of landscape disturbance on mosquito community composition in tropical Australia

2012; Wiley; Volume: 37; Issue: 1 Linguagem: Inglês

10.1111/j.1948-7134.2012.00201.x

1948-7134

Dagmar Meyer Steiger, P. J. Johnson, David W. Hilbert, Scott A Ritchie, Dean A. Jones, Susan G. W. Laurance,

Zoonotic diseases and public health

Emerging infectious diseases are considered to be a growing threat to human and wildlife health. Such diseases might be facilitated by anthropogenic land-use changes that cause novel juxtapositions of different habitats and species and result in new interchanges of vectors, diseases, and hosts. To search for such effects in tropical Australia, we sampled mosquito populations across anthropogenic disturbance gradients of grassland, artificial rainforest edge, and rainforest interior. From >15,000 captured mosquitoes, we identified 26 species and eight genera. Surprisingly, there was no significant difference in community composition or species richness between forest edges and grasslands, but both differed significantly from rainforest interiors. Mosquito species richness was elevated in grasslands relative to the rainforest habitats. Seven species were unique to grasslands and edges, with another 13 found across all habitats. Among the three most abundant species, Culex annulirostris occurred in all habitat types, whereas Verrallina lineata and Cx. pullus were more abundant in forest interiors. Our findings suggest that the creation of anthropogenic grasslands adjacent to rainforests may increase the susceptibility of species in both habitats to transmission of novel diseases via observable changes and mixing of the vector community on rainforest edges. Human-induced land use changes have facilitated the rise and emergence of mosquito-borne diseases worldwide (Walsh et al. 1993, Morse 1995, Reisen 2010). In the tropics, widespread deforestation, road construction, and urbanization have been shown to create habitats that favor arthropod vectors, change host-parasite ecology, and increase disease outbreaks (Patz et al. 2000). Because most emerging diseases in humans are from zoonotic pathogens (transmissible pathogens between animals and humans), this increase in new mosquito-vectored diseases is considered a serious threat to global human health (Taylor et al. 2001, Jones et al. 2008). Although many gaps remain in our knowledge of the dynamics of disease emergence, it is known that environmental changes in deforested landscapes can have a strong influence on vector populations (Tadei et al. 1998, Yasuoka and Levins 2007, Afrane et al. 2008). In the tropics, cleared and forest-edge habitats are often well lit and warmer than rainforest interiors (Camargo and Kapos 1995), characteristics that can increase mosquito larval growth, survivorship, and adult densities (Lindblade et al. 2000, Ye-Ebiyo et al. 2003, Afrane et al. 2006, Vittor et al. 2009). The novel juxtapositions of rainforests with human-altered landscapes can promote new interactions among pathogens, vectors, and hosts. As an apparent result, humans living in deforestation frontiers have suffered heavy exposure to a variety of zoonotic pathogens, such as cutaneous leishmaniasis (Weigle et al. 1993), yellow fever (Barrett and Monath 2003), and cerebral malaria (Rich et al. 2009). In Australia, tropical rainforests occur in parts of far north Queensland although much rainforest has been cleared since European settlement (Keto and Scott 1986, Winter et al. 1987). Prior to European arrival, humans were probably not significant blood sources for mosquitoes as indigenous people were mostly nomadic and few animals were domesticated. The transformation of mosquito domestication most likely occurred with the arrival of Europeans with their livestock and pets (Lee et al. 1980). Although mosquito-borne diseases such as malaria are flourishing worldwide, malaria was declared absent in Australia in 1981. However, northern Australia is still at risk from malaria and other diseases such as dengue and Chikungunya because suitable vectors for these diseases are common and infected people from overseas frequently visit these areas (Longbottom 1996, Hanna et al. 1998). Outbreaks of infections such as dengue, Ross River fever, Barmah Forest virus, Japanese encephalitis, and Murray Valley encephalitis have occurred annually in northern Australia for many years (Mackenzie et al. 1998, Russell and Kay 2004). In urban environments of Australia, mosquitoes are relatively well studied although surprisingly little is known of how land-use changes influence mosquito community composition. A recent study of mosquitoes along an elevational gradient in north Queensland rainforests found them to be most abundant and diverse in the lowlands, where mean annual temperatures are highest (Hilbert 2010). In this study, we focus on the tropical lowlands of north Queensland, contrasting mosquito community composition across gradients of rainforest interiors, rainforest edges, and anthropogenic grasslands in peri-urban settings. We sought to determine whether such habitat characteristics influence mosquito abundance, species richness, and community composition. The study was conducted on the northern outskirts of Cairns city in northeastern Queensland (16° 50′S, 145° 41′E). Annual rainfall is ∼2,000 mm/yr and strongly seasonal with a wet season from December to May. Annual mean monthly temperatures range from 20 to 29° C. Most lowland rainforest in this region was cleared for sugar cane and housing, but the foothills and adjacent mountains still support a closed evergreen forest of 15–26 m in height (Tracey 1982). Forests on the foothills have had a long history of disturbance from selective logging, cyclones, and occasional fires (Webb 1959, Keto and Scott 1986). In our study area, grasslands were human made and dominated by tall (2–4 m high) exotic grasses (e.g., Megathyrus maximus), shrub species, and infrequent pioneer rainforest trees. At each of four study sites (each separated by >700 m), we established three parallel transects per site, with one transect in the forest interior (>100 m from the nearest edge), one along the forest-edge treeline, and one in the adjoining grassland (>100 m from the nearest forest). Adult mosquitoes were collected during the early wet season, in January– February 2011, using CDC (Center for Disease Control) CO2-baited light traps (model 512: John W. Hock Company, U.S.A.). Five traps were placed at 1.5 m heights at ∼20 m intervals along each transect. Traps were operated for 20 consecutive hours, from 15:30 to 11:00 the following day. Because we captured many thousands of mosquitoes, we developed a sub-sampling procedure to randomly sample a subset of captured individuals for identification. All mosquitoes from each trap were distributed on a sheet of graph paper where 30 random points were generated and marked. The mosquitoes on or closest to each point were then identified to species or to genus and morphospecies level. We compared mosquito abundances and species richness across the three habitats types using one-way ANOVAs (with Tukey's post-hoc tests), or with a Kruskal-Wallis Test if data exhibited significant heteroscedasticity. To identify gradients in mosquito communities among habitat types, we used Non-metric Multidimensional Scaling (NMS) with the PC-ORD package (McCune and Mefford 1999). Data were not transformed prior to analysis. Monte Carlo randomizations (100 runs) were used to determine whether the ordination axes explained significantly more variation than expected by chance. Within each habitat type, we assessed the efficacy of our sampling with species-accumulation curves. At the outset, we verified the accuracy of our method for estimating the relative abundance (proportion) of mosquito species, using a large, distinctively yellow species (Coquillettidia near crassipes) that was common in our samples and easily counted. We found strong agreement between its estimated and actual abundance in our samples (rs= 0.750, P<0.0001, Spearman rank correlation; Figure 1), suggesting our subsampling method provided a reasonable approximation of actual species abundances. The positive relationship between actual and subsampled numbers of a distinctive mosquito species (Coquillettidia near crassipes) in our traps suggests that our subsampling technique effectively estimated the relative proportions of species in our samples, especially for more-common species (≥6 individuals). In total, we captured 15,009 mosquitoes in 60 trap-nights across four grassland-edge-rainforest gradients. We averaged 273±151 (mean+SD) mosquitos per trap. Using our subsampling procedure we identified five tribes and two subfamilies of mosquitoes (Table 1). Most (93%) individuals were in the subfamily Culicinae, with only 7% in Anophelinae. Members of the tribe Culicini were found in almost equal proportions across habitats, whereas Aedini were slightly more abundant in forest interiors. Mosquitoes in the Anophelinae subfamily appeared to prefer open habitats, as they were significantly more abundant in grasslands and near edges compared to forest interiors (X2=18.72, P 60% of individuals compared to only 12% for Aedes). Three species constituted 69% of identified captures: Culex annulirostris (44%), Verrallina lineata (14%), and Cx. pullus (11%). C. annulirostris was the most frequently captured species in all habitats with slightly more identified individuals in grassland and edges, whereas Cx. pullus and Ve. lineata were more abundant in forests. Only about half of the identified species were captured in all habitats (Table 2). The remainder were caught in only one or two habitat types. Grasslands and forest edges shared seven species not detected inside rainforests, whereas forest interior and edge sites shared only one species not found in grasslands. Grasslands also had the highest number of unique species followed by forest interiors and then forest edges (Table 2). Habitat type influenced mosquito species richness in an unexpected way (F2,9=8.691, P<0.01; one-way ANOVA; Figure 2). Forest interiors had significantly fewer species than did forest edges (P= 0.013) and grasslands (P= 0.016) (Tukey's tests). Although more mosquitoes were captured in grasslands (38% of all captures) and forest edges (37% of captures) than forest interiors (25% of captures), this difference was not significant because of high among-trap variability in captures within each same habitat type (P=0.040; Kruskal Wallis test). Mean number of mosquito species (±1 SD) captured in each habitat. Mean values in forest interiors were significantly lower than those on forest edges and grasslands. Mosquito community composition varied strongly in response to habitat type. The NMS Axis 1 (which captured 66% of the total variation) and Axis 2 (26% of the variation) both tended to discriminate mosquito communities on forest edges and in grasslands from those in forest interiors (Figure 3). Eight species were significantly correlated with these axes (Table 3), including three associated with forest interiors (Verrallina lineata, Culex pullus, Aedes notoscriptus) and five (Ae. kochi, Ae. lineatopennis, An. bancroftii, An. farauti, Cx. annulirostris) mostly associated with grass and forest edge sites. Ordination of mosquito communities in three different habitats: rainforest interior (filled circles), rainforest edge (filled squares), and anthropogenic grasslands (filled diamonds). Axes 1 and 2 both show gradients between rainforest interiors (clustered in the bottom right-hand side) and rainforest edge and grassland sites (clustered in the top left-hand side). Finally, we assessed patterns of abundance for relatively common mosquito species (≥seven individuals encountered with our subsampling) along the forest interior-grassland gradient. We detected four general patterns (Figure 4): (1) species equally abundant in all habitats; (2) species most common in forest interiors; (3) species common on forest edges; and (4) species most common in grasslands. Thus, the common mosquito species were usually either forest or grassland specialists. Trends in relative abundances of the common mosquito species from forest interior, forest edge, and grassland habitats in tropical Australia. We found several important patterns in our study. First, we found that mosquito communities varied markedly across a rainforest-disturbance gradient. Based on replicated samples at 1.5 m height, rainforest interiors had fewer mosquito species, somewhat fewer total captures, and a different community composition compared to rainforest edges and anthropogenic grasslands. The similarity between rainforest edges and grassland was unexpected because many studies of edge effects on rainforest taxa have shown communities on forest edges to be relatively similar to those in forest interiors (Murcia 1995, Laurance and Bierregaard 1997, Laurance 2004). Although we encountered fewer mosquito species in rainforest interiors than in forest edges or grasslands, we must emphasize that our traps at 1.5 m height would not have encountered all forest-interior species. When we generated species-accumulation curves for each habitat (Figure 5), we found that the curves for grasslands and forest edges appeared to asymptote, whereas the curve for forest interiors was still steadily increasing. This suggests that we failed to encounter and identify a higher proportion of species in rainforest interiors than in the other habitats. Rare species are most likely to be missing from our samples, particularly in rainforest interiors. More extensive sampling that progresses along the wet season may have increased the probability of capturing rare species, particularly those that breed in temporary habitats such as tree hollows (Walker et al. 1991). Species-accumulation curves suggest further sampling would have encountered more mosquito species in forest interiors (A), whereas most of the common mosquito species were sampled in forest edges (B) and grasslands (C). Second, among our common species, we found four general patterns in relative abundance. Only two of thirteen common species were habitat generalists, found in roughly equal abundances across all habitats, whereas most appeared to prefer either grassland or forest habitats. Two species we classified as edge specialists were most likely rainforest-canopy species that had followed the vegetation border down to the ground – a common feature of rainforest-edge communities (Laurance 2004). Determining the basic ecology of mosquito communities is crucial for understanding how deforestation and forest fragmentation influences the vector community and provides insights into where disease risk will increase. From our preliminary study, we suggest that the most probable avenues of novel-disease emergence in changing tropical environments is via mosquito species that are either common across all habitats (e.g., Culex annulirostris) or among species that maintain high abundances up to the margins of forest-grassland edges (e.g., Culex cubicula, Anopheles farauti; Figure 4). Such mosquito species are most likely to be introduced to novel hosts and diseases. Many studies across the tropics have shown that deforestation and other land-uses can lead to an increase in Anopheline and Aedine (Ae. aegypti and Ae. albopictus) mosquitoes (Patz et al. 2000, Matthys et al. 2006, Yasuoka and Levins 2007, Olson et al. 2010). Our study, however, is the first to suggest that grassland mosquitoes can maintain equally high abundances on forest edges. Grassland mosquitoes may be attracted to the shade and humidity on forest edges or are sensitive to prevailing winds (Bidlingmayer and Hem 1981). Other studies have shown that species normally found in the rainforest canopy may descend to the ground near forest edges, a phenomenon we also detected in the two of the species we identified. This has been observed in mosquitoes and other biting flies that vector cutaneous leishmaniasis (Desjeux 2001) and yellow fever in South America (Walsh et al. 1993, Patz et al. 2000). In summary, we found that anthropogenic land-use changes in peri-urban environments in north Queensland strongly influenced mosquito vector communities, which would have potential implications for pathogen transmission to humans and wildlife. The habitat types examined in this study may directly reflect the changes in microclimate and breeding habitat availability to mosquito communities or they may be a proxy for more complex ecological relationships such as host availability and predation and competition with other invertebrates (Lambin et al. 2010). With tropical deforestation continuing apace and strong projected growth in human populations in the tropics, predicting disease risk in these changing environments is an important priority. W.F. Laurance commented on the manuscript. This study was supported by a James Cook University Early Career Grant to S.G. Laurance, MTSRF Transition Funding to D. Hilbert and S. Laurance, and a JCU School of Marine and Tropical Biology grant to D. Meyer Steiger. We thank P. Zborowski for his assistance with difficult identifications.