Understanding how environmental drivers influence the assembly of parasite communities, in addition to how parasites may interact at an infracommunity level, are fundamental requirements for the study of parasite ecology. Knowledge of how parasite communities are assembled will help to predict the risk of parasitism for hosts, and model how parasite communities may change under variable conditions. However, studies frequently rely on presence–absence data and examine multiple host species or sites, metrics which may be too coarse to characterise nuanced within‐host patterns.
We utilised a novel host system, the freshwater mussel Anodonta anatina, to investigate the drivers of community structure and explore parasite interactions. In addition, we aimed to highlight consistencies and inconsistencies between PA and abundance data.
Our analysis incorporated 14 parasite taxa and 720 replicate infracommunities. Using Redundancy Analysis, a joint species distribution model and a Markov random field approach, we modelled the impact of both host‐level and environment‐level characteristics on parasite structure, as well as parasite–parasite correlations after accounting for all other factors. This approach was repeated for both the presence and abundance of all parasites.
We demonstrated that the regional species pool, individual host characteristics (mussel length and gravidity) and predicted parasite–parasite interactions are all important but to varying degrees across parasite species, suggesting that applying generalities to parasite community construction is too simplistic. Furthermore, we showed that PA data fail to capture important density‐dependent effects of parasite load for parasites with high abundance, and in general performs poorly for high‐intensity parasites.
Host and parasite traits, as well as broader environmental factors, all contribute to parasite community structure, emphasising that an integrated approach is required to study community assembly. However, care must be taken with the data used to infer patterns, as presence‐absence data may lead to incorrect ecological inference.