Anthropogenic Landscape Alteration Promotes Higher Disease Risk In ...

Discussion

Our study shows that alteration in the environmental filter (landscape features) leads to increase in infection of Turdus spp. and alteration in the avian species composition, potentially influencing the probability of local avian malaria infection for other species in the community. In addition, we show that climatic variables known to affect the presence and activity of vectors of avian malaria also play a predominant role in shaping the disease transmission structure in New Zealand, with warmer and wetter localities having an enhanced transmission risk of avian malaria.

While the low overall prevalence of infection was unexpected, as many more species are known to be infected by avian malaria (Plasmodium spp.) in New Zealand (see review by [26]), it is not entirely surprising, as true generalist species, such as Turdus species [47], are known to be active reservoirs of vector-borne diseases [2]. The strong phylogenetic signal observed is also in contradiction with what has been observed in other systems, as larger scale studies have shown that Plasmodium spp. usually have a phylogenetic signal ranging from λ = 0.24 to λ = 0.35 [16, 21]. This could be due to the restricted spatial range of our study, encompassing relatively low host diversity, but it could also suggest an increased virulence of the disease in native species at regional scale, causing infected individuals to die or be captured at low rates. In fact, higher virulence of the disease, leading to increased mortality rates in new hosts, has been shown in many other vector-borne disease systems [13]. Increased mortality rates in native New Zealand species, due to enhanced virulence of avian malaria in those species could thus provide a possible explanation for both the low number of infected individuals sampled and the strong phylogenetic signal observed.

Our results suggest that landscape structure drives species community composition at a regional scale, resulting in a higher probability of Turdus spp. presence, potentially influencing local avian malaria occurrence, in accordance with our prediction. We acknowledge that one of our predicted and observed effects suffers from circularity, i.e. the presence of Turdus individuals influences the probability of finding local malaria infections, since only Turdus individuals were found to be infected in our samples. Since many other introduced and native birds in New Zealand are known hosts of malaria (see [26]), we did not expect this limited infection range. Nevertheless, two other observed effects provide insights into avian malaria dynamics: (i) landscape features, i.e. the EVI index, predict the spatial occurrence of Turdus spp., and (ii) climatic variables determine the probability that these reservoirs of malaria are locally infected. While we only included that single bird taxon from our community matrix in larger models, it nevertheless shows that niche availability, dictated by landscape features, directly influences species distribution and subsequent disease risk. Landscape features are known to be a major driver of community composition at regional scales [48]. Indeed, newly created niches in the habitat provide ecological opportunities for new species to colonize those environments [49]. By opening previously pristine habitats and restricted ecological niches, anthropogenic change to the landscape can create spillover of pathogens toward native populations [50], in this case mediated by the intrusion of Turdus reservoirs. For instance, [3] demonstrated that species composition of host assemblages was one of the key drivers of disease transmission success at a local scale for two emerging vector-borne diseases, West-Nile virus and Lyme disease. We believe that this phenomenon provides a suitable explanation in the case of avian malaria in New Zealand avifauna, as introduced species (such as Turdus species) exploit a different niche than native species [51]. Thus, broadening the ecological niche (i.e. higher fragmentation) has probably led to increased interactions and changes in local species pools, enhancing the chance of spillover. Our results also suggest that there is no underlying effect of the landscape on disease transmission once the local pool of species is considered, suggesting that it is not the landscape per se that influences avian malaria, but rather a more complex cascading interaction. Therefore, we believe that our approach allowed us to potentially identify one of the main biotic filters of avian malaria in New Zealand.

A central finding of this study is that both climatic variables used as proxies for the presence of vectors, namely mean annual temperature and mean annual precipitation, have an influence on avian malaria occurrence in the wild. This is in accordance with our prediction, as we expected warmer and wetter parts of the South Island of New Zealand to be more heavily infected due to the increased transmission probability between hosts. This result is also widely supported by the literature, as many other vector-borne diseases using mosquitoes as transmission vectors follow the same pattern [52, 53]. For example, cases of St. Louis encephalitis virus, another mosquito-borne disease, have been shown to be more prevalent during wetter years in the southern United States [54]. The fact that we found that local mean annual temperature influences the probability of avian malaria occurrence could be due to the pathogen maturation time in the vector. From a vector’s perspective, local temperature could play a key role in avian malaria development time in the vector, as maturation times are shorter in warmer conditions [55], thus enabling increased local transmission potential. Moreover, the increase in avian malaria prevalence in wetter and warmer parts of New Zealand could be related to mosquito density, as local climatic conditions are known factors contributing to an increase in mosquito development rates [29, 56], providing a suitable explanation to the pattern we observed. Put in a broader conservation context, as warming events resulting from climate change continue to affect many ecosystems, this could result in an increase in avian malaria into an already disturbed and fragile avifauna, which could potentially lead to local extinction of threatened species.

In conclusion, we show that anthropogenic changes can have dramatic cascading effects on diseases in a wild ecosystem. Of course, there are many limitations to this study which should be considered when interpreting these results. For instance, the limited number of predictors that could be included in the model does not allow complex interactions to be identified; these could arise when considering whole species assemblages instead of only one key species as a proxy of community composition. In addition, we resorted to proxies instead of using the actual variables that may influence avian malaria occurrence (i.e., EVI for landscape features and climatic variables instead of actual mosquito abundance), which limits the inferences that can be made from our results. Moreover, our small number of avian malaria positive individuals, restricted to two species of the same genus, limits our power to make accurate predictions of avian malaria infections across the avian community. Nevertheless, even with our relatively small dataset, we were able to demonstrate clear patterns of avian malaria occurrence in wild New Zealand avifauna. More specifically, we provide evidence of long-lasting effects of anthropogenic changes on the emerging infectious diseases landscape. In the wake of global conservation efforts, we posit that landscape restoration is one of the key factors mediating the impact of emerging infectious diseases in wild ecosystems.