Identifying factors that determine the variation in infection risk in natural populations is of fundamental importance for understanding the ecology and evolution of host-parasite interactions, predicting infection risk and biological conservation [1, 2]. In multi-host, multi-parasite systems, a myriad of factors operating at the individual- and species-level can affect the probability of parasite exposure and subsequent infection across host species [3–6]. At the species level, variation in infection risk can occur because of differences in host life history, behavior and environment that underpin patterns of parasite exposure [7–10]. At the individual host level, hosts can vary in infection risk owing to differences in exposure to parasites and host susceptibility. In the case of vector-borne diseases, hosts exposure to parasites can increase via increase in frequency of encounter with dipteran vectors that can influence disease transmission . For instance, host exposure can be impacted by geographical factors that affect vector abundance (e.g. elevation ) or host ecological traits that affect exposure risk such as foraging/nest height [13, 14] or sociality [10, 15, 16]. Additionally, some species-specific traits associated with disease susceptibility (e.g. sexual dimorphism) or individual-level traits associated with fitness (e.g. fluctuating asymmetry [17, 18] and body condition [19, 20]) could be important predictors of infection risk in natural communities.
Avian haemosporidian parasites (Apicomplexa: Haemosporida) of the genera Plasmodium, Haemoproteus (including Parahaemoproteus) are protozoan blood parasites that affect bird populations globally . Avian haemosporidians (commonly referred to as avian malaria parasites) are an exceptionally diverse group of parasites, with over 2500 parasite genetic lineages . These parasites are transmitted by arthropod vectors, with Plasmodium being transmitted by culicid mosquitoes, and Haemoproteus by ceratopogonid biting midges and hippoboscid louse flies [21, 23]. Avian haemosporidians can impose strong selective pressures on bird hosts as they are known to reduce longevity , host fitness [25, 26], individual host condition  and have led to severe population declines [28–31]. Previous research has revealed that avian haemosporidian parasites vary widely in their host range, with Plasmodium lineages often being generalists infecting a broad range of host species and Haemoproteus lineages often being specialists infecting one or few closely related host species [32–34] but this pattern is not universal [35, 36]. Plasmodium and Haemoproteus parasites are also known to exhibit eco-evolutionary differences, with community structure of Plasmodium generally affected by abiotic factors such as spatial proximity and Haemoproteus, primarily affected by biotic factors such as host phylogeny and host ecology [32, 37, 38] but see [39, 40]. Given the widespread distribution, diversity and pronounced eco-evolutionary differences between Plasmodium and Haemoproteus, variation in parasite prevalence for the two parasite genera could be affected by different ecological and evolutionary factors in multi-host communities.
Several studies have attempted to identify ecological factors that can predict haemosporidian infection risk in avian communities, but offer mixed support, in part, owing to the limited exploration of concomitant factors simultaneously across entire host assemblages and across both parasite genera or challenges associated with understanding complex interactions operating across different scales (e.g. within and between species). For instance, specific habitat and temperature requirements of different haemosporidian vectors (e.g. mosquitoes, biting midges) and parasites may limit their distribution on an elevational gradient and across habitat types [12, 41]. While some studies support higher prevalence of Haemoproteus at high elevations and high prevalence of Plasmodium at lower elevations [42, 43] others present contrasting patterns, with high prevalence of Haemoproteus and Plasmodium at mid-elevations  and no effect of elevation for Haemoproteus and Plasmodium parasites .
Roosting and foraging stratum of host species may increase the probability of hosts encountering vectors thereby promoting parasite transmission. It has been hypothesized that social aggregations attract vectors, and this could lead to higher prevalence of avian haemosporidians in social species [16, 45]. Furthermore, vertical stratification in arthropod vectors could result in variation in avian haemosporidian infection risk due to differences in abundance of vectors in the canopy compared to the ground level [46–48]. Other species ecological traits such as host specialization (e.g. habitat or elevational specialization) could also lead to differential exposure to vectors/parasites. Generalists species, spanning a wider elevational range, may encounter more vectors or a greater diversity of habitats compared to elevational specialists, leading to higher parasite prevalence in generalists than specialists. Furthermore, previous studies suggest host species that exhibit sexually dimorphic traits (e.g. plumage brightness) have likely been exposed to higher levels of parasitism because parasite pressure is a strong driver of sexual selection on these traits [49, 50]. Thus, species that are more susceptible to parasite infections likely exhibit higher levels of sexual dimorphism .
At the individual level, previous studies suggest that birds with higher average body size tend to have higher infection probability as larger body size will likely provide more surface area for vector feeding and emit higher quantity of olfactory cues (e.g. CO2), thereby attracting more vectors [14, 51]. Host body condition can also affect the likelihood of infection due to differences in individual susceptibility. A negative association between body condition and parasitism is generally expected, either due to reduced immunocompetence in birds with poor body condition and increased susceptibility or the direct effects of parasitism on the fitness of individuals, leading to poor body condition [4, 20]. Furthermore, fluctuating asymmetry (defined as small, random deviations from symmetry of bilateral symmetrical traits ) could be an important predictor for avian haemosporidian infection. The positive association between fluctuating asymmetry and parasitism is relatively common in natural populations [17, 18], and this association could exist either because parasitism (as a form of developmental stress) can directly increase levels of fluctuating asymmetry, or because individuals with high fluctuating asymmetry have low immunity.
Although several factors have been proposed to explain variation in parasite prevalence and infection risk among individuals and host species [13, 14, 43, 48, 51, 53], it remains unclear whether the role of host ecological traits are generally predictable or whether they are idiosyncratic across hosts, parasites and environmental conditions and context-dependent. Additionally, evolutionary history of host species can confound the relationship between ecological traits and parasite infection risk as closely related species are more likely to share risk factors compared to non-related host species. Despite this, surprisingly few studies have taken evolutionary history of the hosts into account (e.g. [48, 54]) and thus, the importance of host evolutionary history in predicting infection risk is less well understood.
The Tropical Sky Island bird community in the Western Ghats mountains, located parallel to the southern coast of India (Fig. 1), offers an excellent model system to elucidate the factors influencing variation in avian haemosporidian infection risk. The Western Ghats are a global biodiversity hotspot , and the high endemic bird diversity in the Western Ghats  provides opportunities for native parasites to exploit a wide variety of hosts, allowing us to test how host ecology impacts parasite infection risk. Additionally, the landscape is threatened by anthropogenic habitat fragmentation and land-use changes; and the potential negative impact of avian malaria in this biodiversity hotspot makes the identification of factors associated with increased infection risk an important step for conservation .
Sky Islands are isolated mountain-top habitats surrounded by dramatically different lowland habitats. The replicated arrangement of geographically discrete, identical habitats provides an ideal natural laboratory to explore ecological dynamics underlying avian haemosporidian infection risk. The Western Ghats mountains are interrupted by three major bio-geographical breaks, the Chaliyar River valley (2–3 km wide), the deepest Palghat Gap (40 km wide) and the Shencottah Gap (10 km wide), resulting in genetic differentiation and speciation across a range of taxa [58–60]. Such patterns of genetic differentiation in hosts could impact the spatial distribution of parasite populations harboring these bird hosts. Within each Western Ghats mountain, Sky Islands hosts unique natural matrix of wet, montane evergreen forests and grasslands, locally known as Sholas, above 1400 m (henceforth Shola Sky Islands), while low elevations harbor drier habitats. High habitat heterogeneity and climatic conditions due to its elevational gradient have led to disproportionately high host species diversity in the Shola Sky Islands, comprising of host species having different habitat specialization, life history strategies and elevational distribution. For example, montane specialists are restricted to high elevations and generalists are distributed widely from high to low elevations. While montane specialists have likely been historically protected from avian malaria because low temperatures at high elevations leads to low vector abundance  or poor parasite development , this scenario might be changing as global warming progresses [41, 62]. Thus, Western Ghats Sky Islands offer a valuable system in which to investigate infection dynamics, especially in the light of possible climate change driven extinctions in the landscape (e.g. Robin et al. ).
In this study, we first examine the species- and individual-level ecological factors that influence variation in avian haemosporidian prevalence and thus avian haemosporidian infection risk in the Western Ghats. Next, we examine if these effects differ across the two parasite genera, Plasmodium and Haemoproteus. Secondly, we test the effect of host evolutionary history in explaining variation in avian haemosporidian infection risk not explained by host ecological factors. As mentioned earlier, several studies suggest that Plasmodium is a generalist parasite and Haemoproteus is a relatively specialist parasite [32–34], thus we expect that the effects of ecological factors will vary for Plasmodium and Haemoproteus in addition to their intrinsic differences in parasite biology and vector specificity. At the species-level, we expect that: (i) species that have a lower minimum elevation will have higher Plasmodium prevalence whereas species with a higher minimum elevation will have higher Haemoproteus prevalence (Fig. 2); (ii) species foraging at higher forest strata will have lower Plasmodium prevalence and higher Haemoproteus prevalence compared to species foraging at the ground level (Fig. 2); (iii) social living species and species with sexually dimorphic traits will likely exhibit higher parasite prevalence of both parasites (Fig. 2). Furthermore, at the individual-level, we expect: (iv) infection risk would increase with increase in body size and fluctuating asymmetry (Fig. 2); and (v) birds with better body condition will likely be less infected by both parasites compared to birds with poorer body condition (Fig. 2).