Long‐term capture data uncover shifts in the daily activity patterns of Amazonian birds

Although the duration of biological rhythms varies from milliseconds to decades, daily cycles are especially widespread for individual organisms. These circadian rhythms are synchronized by light–dark cycles, which govern predictable changes in other microclimate variables. Despite the fixed, regular nature of these abiotic cycles at low latitudes, little is known about how different animals structure activity along these gradients. We used 25 years of capture–recapture data and ~ 25 000 unique captures to characterize ‘activity' (time of capture) for 65 species of Amazonian birds in the lowest forest stratum (0–2.5 m). To quantify vertical niche breadths, we also measured foraging heights for these same species in the field. Activity peaked within the first two hours after sunrise for most species (71%), while others peaked 1–3 h later (29%). This difference was associated with an apparent preference for specific vertical strata, as terrestrial and near‐ground birds peaked ~ 1 h after sunrise, while birds from higher strata delayed their first bouts of activity in the lower understory (by an average of 40–100 min). Peak capture times also correlated with long‐term abundance trends, as only species with the earliest capture times showed strong declines, suggesting that associations with mid‐ or late morning microclimates may contribute to resiliency in birds typical of higher strata. Because arboreal species largely avoid the lower understory in the early morning, when terrestrial birds are active, we infer that they are poorly adapted to low light environments. However, with their broad vertical niches and reliance on brighter microhabitats, we suggest that an arboreal lifestyle diminishes specialization and may increase the capacity to cope with a wider range of daily microclimate conditions. We predict that subtle, but important, shifts in daily activity patterns would materialize if other biodiversity‐monitoring programs could analyze similar temporal datasets.