Spatiotemporal variation in drivers of parasitism in a wild wood mouse population

Host–parasite interactions in nature are driven by a range of factors across several ecological scales, so observed relationships are often context‐dependent. Importantly, if these factors vary across space and time, practical sampling limitations can limit or bias inferences, and the relative importance of different drivers can be hard to discern. Here we ask to what degree environmental, host and within‐host influences on parasitism are shaped by spatiotemporal variation. We used a replicated, longitudinal dataset of >1,500 observations for nearly 1000 individual wood mice, Apodemus sylvaticus, encompassing 6 years of sampling across five different woodland sites, and investigated drivers of infection intensity with a highly prevalent gastrointestinal nematode, Heligmosomoides polygyrus. We used a Bayesian modelling approach to further quantify if and how each factor varied in space and time. Finally, we examined the extent to which a lack of spatial or temporal replication (i.e. within single years or single sites) would affect which drivers predict H. polygyrus infection. Season, host body condition, and sex were the three most important determinants of infection intensity; however, the strength and even direction of these effects varied in time, but not in space. Models fit to single years and site replicates in many cases showed weak and variable detection of effects of the factors investigated, highlighting the benefits of long‐term sampling for separating meaningful ecological variation from sampling variation. These results highlight the importance of accounting for spatiotemporal variation in determining what drives disease dynamics and the need to incorporate replication in both time and space when designing sampling regimes. Furthermore, we suggest that embracing, rather than simply controlling for, spatiotemporal variation can reveal meaningful variation for understanding the factors impacting parasitism (e.g. season and host charateristics) which can improve predictions of how wildlife health will respond to change. A free Plain Language Summary can be found within the Supporting Information of this article.