key: cord-1026983-fhlff0va
authors: Okanga, Sharon; Cumming, Graeme S.; Hockey, Phillip A. R.; Peters, Jeffrey L.
title: Landscape structure influences avian malaria ecology in the Western Cape, South Africa
date: 2013-10-08
journal: Landsc Ecol
DOI: 10.1007/s10980-013-9949-y
sha: 831b75602a37495a0fc4968a1dd9db50590f975d
doc_id: 1026983
cord_uid: fhlff0va

A central theme in landscape ecology is that of understanding the consequences of landscape heterogeneity for ecological processes. The effects of landscape heterogeneity on parasite communities are poorly understood, although it has been shown that anthropogenic impacts may contribute to outbreaks of both parasites and pathogens. We tested for effects of landcover type, composition, configuration, and urbanisation on avian diversity and avian malaria prevalence in 26 communities of wetland-associated passerines in the Western Cape of South Africa. We predicted that avian malaria prevalence would be influenced by the pattern of farmland and urban areas in the surrounding landscapes and the sizes of the wetlands in which birds were sampled. We quantified landscape pattern using a six-class simplification of the National Landcover data set at 35 × 35 m resolution and five extents of between 1 and 20 km from each wetland. The bird community was sampled using point counts and we collected blood samples from birds at each site. We screened these for malaria using PCR and molecular techniques. Passerine species richness and infection prevalence varied significantly between different landcover types. Host richness and parasite prevalence were highest in viticultural and cropping sites respectively and lowest in urban sites. Wetlands located in indigenous vegetation had intermediate numbers of bird species and intermediate parasite prevalence. Landscape composition and habitat type surrounding wetlands emerged as useful correlates of infection prevalence. Anthropogenic landscape modification appears to have both direct and indirect effects on avian communities and their associated parasite assemblages, with attendant consequences for avian health.

The consequences of landscape heterogeneity for avian populations and communities have been extensively studied. Far less is known about the consequences of landscape heterogeneity, and of changes in landscape composition and configuration, for parasites and pathogens. As humans alter landscape pattern, a number of impacts on parasites and their hosts and vectors are likely to occur. Of particular relevance are changes in the abundance and species richness of host and vector species. These occur in response to changes in such variables as habitat structure, resource availability, predation rates, and ease of dispersal (With and Crist 1995; Flather and Sauer 1996) . Unless all hosts and all vectors are equally competent, changes in host and/or vector communities can be expected to alter the transmission rates of parasites (and their overall prevalence within the host community) by altering the numbers of suitable and available hosts (Kruess and Tscharnkte 1994; Brisson et al. 2011) , the number of successful transmission events that occur by either direct contact or via vectors, and/or the prevalence of infectious individuals among susceptible hosts (Roche et al. 2012) .

The problem of deriving general principles from which to predict impacts of landscape change on parasite communities is complicated by a set of indirect and idiosyncratic effects. For example, more diverse host communities may exhibit lower infection rates because more 'wasted bites' on incompetent vector species occur (Dobson et al. 2006) . Conversely, communities that are more diverse to begin with will be more likely to retain at least some highly competent hosts and vectors following anthropogenic habitat modification. Changes in landscape composition and configuration can influence dispersal rates and bring domestic and wild species (and people) into contact with one another, further altering parasite and pathogen dynamics. For instance, deforestation has been blamed for outbreaks of human malaria in the Amazon Basin (Vittor et al. 2006) , cutaneous leishmaniasis in Costa Rica (Chaves et al. 2008) , and severe acute respiratory syndrome (SARS) in South East Asia (Field 2009 ). Urbanization has similarly been linked to the proliferation of zoonotics such as Lyme disease (LoGiudice et al. 2003 ) and rabies (Singh et al. 2001) .

Although birds are highly mobile, they may have highly specific resource requirements. In southern Africa, Child et al. (2009) identified a decline in species richness at upper trophic levels between protected (and mostly intact) and non-protected (fragmented) landscapes. Similarly, urbanization can have a negative effect on habitat quality and correspondingly on avian species richness in greater Cape Town (Dures and Cumming 2010) . These changes would be expected to both directly and indirectly affect parasite and pathogen prevalence. In the USA, research on West Nile Virus (WNV) showed that an increased prevalence of antibodies for WNV in songbirds corresponded to an increasing scale of urbanization (Bradley et al. 2008) . In American wetlands, WNV prevalence decreased with increasing wetland size (Ezenwa et al. 2007) . Similarly, wetland size has been linked to resource availability (Paracuellos and Telléria 2004) , and may subsequently affect local avian species richness as a result.

We tested three predictions about the ecology of avian haemosporidia in the Western Cape, South Africa, with a focus on wetland passerine communities. First, we predicted that infection prevalence would be influenced by the composition of the surrounding landscape, and especially of urban areas, since these adversely impact bird species diversity and richness (Fair child et al. 2009; Dures and Cumming 2010) . Species diversity and species richness have both been linked to infection prevalence Swaddle and Calos 2008) . Second, we predicted that infection prevalence would be affected by the perimeter-area ratio and connectivity of habitats, as these potentially influence encounter rates between pathogens and hosts (as described above from other disease systems; viz Singh et al. 2001; Vittor et al. 2006; Chaves et al. 2008) . Third, we predicted that wetland size would affect infection prevalence, because of its impacts on resource availability and species richness.

Twenty-six wetlands were selected as study sites from within the historical boundaries of the fynbos biome in the Western Cape. The area has strong gradients of human population density, ranging from a large city (Cape Town) to sparsely settled agricultural and rural areas; a diversity of land uses, ranging from intense crop agriculture, viticulture and forestry to virgin protected areas; strong gradients of rainfall, temperature and altitude; and indigenous vegetation, the fynbos, that is mostly either shrubland or dwarf shrubland. Sites were situated along an urban to rural gradient and restricted to altitudes of 0-300 m above sea level (Fig. 1) . Each site was visited twice over the course of 2 years (2010 and 2011), with samples collected once during the summer (January to March) and once during the winter (July to September). Wetland size was determined by mapping the perimeter of the main water body at each site using a handheld GPS unit (Garmin GPS map 76CSx).

Birds were trapped using mist nets, with efforts focused on wetland passerines. Blood samples were taken by pricking the brachial vein with a 26-gauge needle (26G 9 inch) and capturing 30-50 ll of blood in a capillary tube (Dawson 2004) . The blood was transferred into a vial containing lysis buffer. The vial was sealed and stored and all birds were released after sampling. Five birds were recaptured during the sampling period; these individuals were counted once during analysis.

Blood samples were analysed following a PCR protocol, as detailed by Cumming et al. (2013) . Two PCR trials were conducted for each sample under identical conditions. PCR products were sequenced using a 10 ll reaction using the Big Dye Terminator v.3.1 Cycle Sequencing Kit (Applied Biosystems), following manufacturer protocols. The final products were sent for automated sequencing at the DNA Sequencing Facility on Science Hill, Yale University. This PCR routine only detected parasites within the Plasmodium and Haemoproteus genera (Waldenström et al. 2004) .

Spatial analysis was conducted using ArcGIS (9.3.1_2008) and Fragstats (version 4) (McGarigal et al. 2012) . Additional statistical analysis was conducted in R (2011_12_22). Analysis of the landscape was approached with two objectives: (1) to determine the influence of landscape composition (areas of different landcover types) on bird communities and prevalence of avian haemosporidia; and (2) to determine the influence of landscape configuration (spatial arrangement, including perimeter to area ratios, edges, and connectivity) at various extents.

To classify landcover we used the South African National Landcover Data set (NLCD), which is derived from interpretation of Landsat 7 TM imagery at 35 m resolution (Fairbanks et al. 2000) . We reclassified the NLCD to yield a simpler subset of the most relevant land cover types: fynbos, agriculture, urban, and viticulture. Sites were then grouped according to the dominant surrounding land cover class: fynbos (3 sites), agricultural (9 sites), urban (7 sites) and viticultural (7 sites). Other nearby landcover types (mining, forest, etc.) were reclassified into a single category labelled 'other' (Fig. 1) .

Buffer zones of 1, 5, 10, 15 and 20 km in extent were created around each wetland in ArcGIS and extracted to give corresponding landscape patches. These extents were chosen to correspond to the range of reported flight distances (scales of landscape use) of birds from the family Ploceidae (bishops, weavers, and allies), which can be up to 20 km in weaver birds (Hockey et al. 2005) . We based the analysis on the Ploceidae because our data show that they were the family that was most prevalently infected with avian malaria and we expected that their incidence and movements would relate closely to infection patterns.

Variables specific to landscape area, edge, and connectivity were quantified using Fragstats (McGarigal et al. 2012) , at each of our five extents of analysis, to determine which structural aspects were influential to infection prevalence (see Table 1 ).

Prevalence data for Plasmodium and Haemoproteus were combined and referred to as infection prevalence. Bird species richness (site species richness, infected species richness and passerine species richness) was taken into account due to its response to landscape structure (Child et al. 2009 ).

We used one-way ANOVA and parametric regression analysis to determine the nature of the relationships between landscape type and pattern; wetland size; urban density; infection prevalence and bird species richness (site and passerine). Due to the large number of comparisons being made, a Bonferroni correction was made for critical p values at the 0.05 and 0.01 level, to reduce the probability of type I errors (Dunn 1961) . Critical p values were set as

where Bfc is the Bonferroni correction factor, the subscript indicates the probability value to which the Bfc applies, and 26 was the number of comparisons made. TCA total core area (m 2 ) of each landcover type, total core area, P total perimeter, A area, N c number of joins with patch of same landcover type, T c total connections possible (of same landcover type), Np number of landcover types, Tp total number of landcover types, wa wetland area (m 2 ), nb number of buildings within a 1 km radius of wetland, ws wetland size Generalized linear mixed models (GLMMs) were used to explore which landscape parameters best explained infection prevalence patterns. Models were fitted with canonical links (corresponding to the nature of the prevalence distribution), with data arranged by extent. Area, connectivity, perimeter-area ratio (PARA) and landcover type featured as fixed effects in models, and passerine species richness and site featured as factors. We ranked candidate models according to their Akaike's Information Criterion (AIC) value and retained or excluded variables according to their effect on the AIC value (Burnham and Anderson 2002) , with lower AIC values indicating a better fit of a model to the data.

A total of 200 out of 974 samples were both positive for avian malaria and successfully sequenced, with the majority of infected species belonging to the family Ploceidae (n = 165). The majority of birds were sampled from wetlands within agricultural landscapes (344), followed by wetlands within viticultural landscapes (308); then fynbos and urban landscapes (98 and 224 birds respectively). Sites within a viticultural landscape had the highest mean species richness (all birds) whereas agricultural sites had the highest mean passerine species richness. Fynbos sites had the lowest mean site species richness (Fig. 2) . There was little difference in mean infected species richness, which ranged from 9 to 12 between landscapes. Mean infected species was highest in viticultural landscapes (12) and lowest in urban landscapes (9). Urban landscapes also had the lowest mean passerine species richness (16). Significant variation in infection prevalence within different landscapes did not occur with either site species richness (F 3, 22 = 0.96; p = 0.43) or infected species richness (F 3, 22 = 1.16; p = 0.35); however, passerine species richness was associated with landscape type (F 3, 22 = 3.55; p = 0.03).

There was a clear disparity in the mean infection prevalence of birds sampled from wetlands within different landscapes (F 3, 22 ; p = 0.002). Birds sampled from agricultural landscapes had the highest infection prevalence, whereas birds sampled from urban landscapes had the lowest overall infection prevalence (Fig. 3) . This pattern was also apparent when the regional landscape was viewed as a whole (Fig. 4) .

After Bonferroni correction, urban landcover stood out as the most influential landcover type on passerine species richness (Table 2) . Increasing the area of urban landcover had a consistently negative influence on passerine species richness, with the largest effects seen at the 15-20 km extents. Passerine species richness declined with increasing urban density (r 2 = -0.40; p = 0.01). Edge and connectivity of urban cover was also significantly related to passerine species richness, although only at the 1 km extent. Area and connectivity of agricultural land was similarly influential on infection prevalence and species richness at the more local extents (1-5 km). Also notable was the influence of edge of other landcover types on infection prevalence at the 20 km extent.

Variation in species richness and prevalence with landcover structure Urban proximity was also influential on infection prevalence, with increased proximity causing a decline in prevalence (r 2 = 0.40; p = 0.04). Other species richness measures did not show any notable association with either building density or urban proximity; site species richness was, however, significantly associated with wetland size (r 2 = 0.43; p B 0.001).

Using generalized linear mixed modelling, the best fitting models featured site as a highly influential factor with landcover type or passerine species richness as mixed effects (Table 3) . Site, landcover type and passerine species richness were the most influential factors to infection prevalence, with negligible variation between models featuring only these variables (DAIC B2). Structural parameters (area, perimeter-area ratio and connectivity) produced a poorer model fit and were comparable to each other in terms of their influence, with no discernible difference in fit seen when one parameter was substituted for another.

In general, wetlands within agricultural and viticultural landscapes ranked highest for passerine and site species richness respectively. By contrast, infected species and passerine species richness were lowest in fynbos and urban landscapes (Fig. 2) . This outcome seemed to reflect habitat preferences of wetland passerines. The heaviest malarial infections and highest passerine species richness were both found in agricultural landscapes, whereas the lowest values occurred in urban landscapes (Figs. 2, 3) . Birds from the Ploceidae family also exhibited some of the heaviest infection prevalences. Many sampled passerine species were seed-eaters (Hockey et al. 2005) , occurring mostly in the Ploceidae and Passeridae (sparrows and allies). It is natural that seed-eaters would be commoner in agricultural landscapes, as these landscapes are presumably richer in available resources for seed-eating birds. The finding also concurs with previous studies, where bird species richness mirrored resource availability (Paracuellos and Telléria 2004; Child et al. 2009 ). 

GLMMs showed that landcover type, site and passerine species richness were the variables best describing infection prevalence (Table 3) . Model fit changed only minimally when configurational variables were added, indicating that landscape configuration was not as influential as landcover type. This result is comparable to the outcomes from related studies on disease prevalence in various landscapes, where infection prevalence varied significantly between different vegetation types (Boone et al. 2000) , and to more general studies that have indicated a larger effect of habitat amount than of habitat configuration (e.g., Flather and Bevers 2002) . Other studies on vectorborne disease prevalence and landscape composition and configuration have, however, indicated that both land use and connectivity can also play primary roles in influencing disease prevalence (Linard et al. 2007; Vanwambeke et al. 2010) . These outcomes suggest that host and vector ecology not only contribute to the influence of landscape factors in vector-borne disease, but also affect the magnitude of their influence. Wetlands in agricultural landscapes had the highest infection prevalences among those sampled (Fig. 3) . While this could be solely a consequence of the fact that passerine (host) species richness was highest in agricultural areas, vector-related factors may also have played a part in this outcome. Previous studies in Kenya and Ethiopia have demonstrated a positive link between Anopheles gambiae larval development and the agricultural modification of landscapes, which often enhances conditions for mosquito breeding (Kebede et al. 2005; Munga et al. 2009 ). It is feasible that the same mechanisms may be in effect in the Western Cape of South Africa, where crops can provide a rich nutrient source for the larvae of avian malaria vectors. Additionally, irrigation practices in agricultural landscapes in the Western Cape (Venter 2005) facilitate vector breeding and development, as virtually all avian malaria vectors have a waterdependent stage in their lifecycle (Valkiūnas 2005) .

Wetland size, landscape configuration and composition were not associated with each other, as wetlands of assorted sizes occurred in a variety of landscapes. Previously, wetland area has been shown to affect both pathogen prevalence and avian species richness (Paracuellos and Telléria 2004; Ezenwa et al. 2007 ). In our study, wetland size did not exhibit any apparent association with infection prevalence. Wetland size, however, was significantly associated with site species richness, suggesting that it exerts an indirect influence on passerine species richness.

The urban landscape, bird species richness and infection Urban landcover was consistently influential on passerine species richness; urban edge and connectivity in particular both exerted a direct influence on passerine species richness, with the greatest effect observed at the most local extents. The largest influence of landscape structural factors was at local extents (1-5 km) and restricted to specific landscape types (agricultural and urban). This was also an indication that factors pertaining to species richness with the most influence on prevalence patterns were most intense at close proximity ( Table 2) . The low passerine species richness and correspondingly low infection prevalence in urban areas (Figs. 2, 3) suggested that most sampled passerine species either do not adapt Bradley et al. 2008 ). An alternative explanation is that the encounter rate in urban landscapes between passerine hosts and malarial pathogens is much lower than in other landscapes due to a lack of infectious vectors and/or susceptible hosts (Roche et al. 2012) .

Host adaptability to the landscape can have similarly significant implications for host viability and disease prevalence. Piersma (1997) postulated that some avian species are restricted to parasite-poor habitats as a trade-off with investing in immunocompetence, resulting in some species spending much more energy migrating to parasite-poor habitats in order to avoid infection. This hypothesis prompts several considerations: for instance, are migratory birds less prevalently infected than resident species? Do urban landscapes qualify as 'parasite-poor' habitats in the avian malaria model? There is also the possibility that birds act as sources of disease, as was the case in Chicago, USA, where migratory birds were the main hosts promoting the invasion of ticks (including nonendemic species) into the city (Hamer et al. 2012 ). Patterns of disease across a landscape are often the consequence of spatially explicit and interlinked factors that respond similarly to changes in landscape structure (Patz et al. 2000) . The response of pathogens and hosts to changes in the landscape may vary widely. Our findings contrast with those from a West African study on avian malaria, which reported a higher prevalence of avian malaria in undisturbed habitats compared to modified (deforested) habitats (Chasar et al. 2009 ). In our study, undisturbed habitats (fynbos) had lower infection prevalence than agriculturally modified habitats (agricultural and viticultural). Our findings also contrast with several other related studies, where the prevalence of a variety of diseases in wildlife populations increased with, or was promoted by, urban land development (Singh et al. 2001; Farnsworth et al. 2005 ). On the other hand, a human malaria study found a lower prevalence in urban areas, which concurred with outcomes seen in this study (Omumbo et al. 2005) . Our overall impression is therefore that landcover can act as either a positive and negative influence on disease, with its exact role being co-dependent upon a diversity of mechanisms in host and vector ecology (Patz et al. 2000) .

Land use and avian species diversity

Remotes sensing and geographic information systems: charting sin nombre virus infections in deer mice

Urban landuse predicts West Nile virus exposure in songbirds

It takes a community to raise the prevalence of a zoonotic pathogen

Model selection and multimodel inference: a practical information-theoretic approach

Prevalence and diversity patterns of avian blood parasites in degraded African rainforest habitats

Social exclusion modifies climate and deforestation impacts on a vector-borne disease

Assessing the broadscale impact of agriculturally transformed and protected area landscapes on avian taxonomic and functional richness

Host associations, biogeography, and phylogenetics of avian malaria in southern African waterfowl

Techniques in physiology and genetics: blood sampling

Sacred cows and sympathetic squirrels: the importance of biological diversity to human health

Multiple comparisons among means

The confounding influence of homogenising invasive species in a globally endangered and largely urban biome: does habitat quality dominate avian biodiversity?

Landcover variation and West Nile Virus prevalence: patterns, processes and implications for disease control

The South African landcover characteristics database: a synopsis of the landscape

Human land use influences chronic wasting disease prevalence in mule deer

Bats and emerging zoonoses: henipaviruses and SARS

Patchy reaction-diffusion and population abundance: the relative importance of habitat amount and arrangement

Using landscape ecology to test hypotheses about large-scale abundance patterns in migratory birds

Wild birds and urban ecology of ticks and tick-borne pathogens

Roberts' Birds of Southern Africa, 7th edn. The trustees of the John Voelcker bird book fund, Cape Town Kebede A

Effects of species diversity on disease risk

Habitat fragmentation, species loss and biological control

Environmental conditions and Puumala virus transmission in Belgium

The ecology of infectious disease: effects of host diversity and community composition on Lyme disease risk

FRAGSTATS v4: spatial pattern analysis program for categorical and continuous maps. Computer software program produced by the authors at the University of Massachusetts

Urbanization as a major cause of biotic homogenization

Land use and landcover changes and spatiotemporal dynamics of Anopheline larval habitats during a four-year period in a highland community of Africa

The influence of urbanisation on measures of Plasmodium falciparum infection prevalence in East Africa

Factors affecting the distribution of a waterbird community: the role of habitat configuration and bird abundance

Effects of environmental change on emerging parasitic diseases

Do global patterns of habitat use and migration strategics co-evolve with relative investments in immunocompetence due to spatial variation in parasite pressure

Linking community and disease ecology: the impact of biodiversity on pathogen transmission

Epidemiological characteristics of rabies in Delhi and surrounding areas

Increased avian diversity is associated with lower incidence of human West Nile infection: observation of the dilution effect

Avian malaria parasites and other haemosporidia

Landscape predictors of tick-borne encephalitis in Latvia: landcover, land use and land ownership

FAO report: fertilizer use by crop in South Africa. Land and Plant Nutrition Management Service, Land and Water Development Division

The effect of deforestation on the human-biting rate of Anopheles darlingi, the primary vector of falciparum malaria in the Peruvian Amazon

Critical thresholds in species' responses to landscape structure

Acknowledgments Many thanks to David Nkosi for his assistance with fieldwork. We are grateful to private landowners in the Western Cape, the City of Cape Town, SANParks and Cape Nature, for granting access to sites in urban and rural sites and within reserves. This work was funded by grants from the DST/NRF Centre of Excellence at the Percy FitzPatrick Institute of African Ornithology and the International Foundation for Science (IFS).