Avian malaria (Plasmodium spp.) and other blood parasitic infections of birds constitute increasingly popular model systems in ecological and evolutionary host–parasite studies. Field studies of these parasites commonly use two traits in hypothesis testing: infection status (or prevalence at the population level) and parasitaemia, yet the causes of variation in these traits remain poorly understood. Here, we use quantitative PCR to investigate fine-scale environmental and host predictors of malaria infection status and parasitaemia in a large 4-year data set from a well-characterized population of blue tits (Cyanistes caeruleus). We also examine the temporal dynamics of both traits within individuals. Both infection status and parasitaemia showed marked temporal and spatial variation within this population. However, spatiotemporal patterns of prevalence and parasitaemia were non-parallel, suggesting that different biological processes underpin variation in these two traits at this scale. Infection probability and parasitaemia both increased with host age, and parasitaemia was higher in individuals investing more in reproduction (those with larger clutch sizes). Several local environmental characteristics predicted parasitaemia, including food availability, altitude, and distance from the woodland edge. Although infection status and parasitaemia were somewhat repeatable within individuals, infections were clearly dynamic: patent infections frequently disappeared from the bloodstream, with up to 26% being lost between years, and parasitaemia also fluctuated within individuals across years in a pattern that mirrored annual population-level changes. Overall, these findings highlight the ecological complexity of avian malaria infections in natural populations, while providing valuable insight into the fundamental biology of this system that will increase its utility as a model host–parasite system.