Metacommunity theory provides a framework to understand how ecological communities vary in space and time. However, few studies have investigated metacommunity processes in a context of long term changes. Environmental changes can impact species distribution and therefore the structure of metacommunities. Using two complementary methods, this study evaluated the temporal variability of bird metacommunity processes in an agricultural landscape after 25 years of changes in land-cover. Bird and landscape data were recorded in the same locations using a series of 256 point counts in 1982 and 2007. First, variance partitioning was applied to quantify the roles of environmental filtering (i.e. landscape composition variables) and spatial variables to organize bird metacommunities each year. Second, changes in the structure of the bird metacommunity were examined by quantifying three of its components: coherence, species turnover and species range boundary clumping. Our results demonstrate that landscape variables explained slightly more bird metacommunity patterns than spatial variation of unknown origin each year. The bird metacommunity had a Clementsian structure (i.e. grouped distribution of species along environmental gradients) which was correlated with a landscape gradient ranging from open farmland to wooded sites. This structure was similar each year. To conclude, the study shows that environmental filtering with specializations to different habitats is a major process in determining bird metacommunities in landscapes. Moreover, our results suggest that metacommunity structure can remain constant over time despite demographic and environmental changes