Habitat heterogeneity and demographic processes create variability in the major taxonomic diversity trends: 1) biotic homogenization and 2) the emergence of novel community compositions. Nonetheless, little is known about how the imprints of environmental filtering and random demographic processes on community dissimilarity vary over 1) time or 2) space. Quantifying such variation is key to revealing temporal regime shifts, latitudinal trends, and site-level specificity in the drivers of community dissimilarity.

To characterise variation in drivers of community change, we introduce the concept of ‘non-stationary community responses'. We then apply this concept to estimate temporal and spatial variability in the imprints of climate, land cover and random processes on spatial and temporal dissimilarity of community composition. As a model system, we use multidecadal monitoring data of bird (1147 monitoring sites; 49 years), butterfly (101 monitoring sites; 22 years), and moth (99 monitoring sites; 26 years) communities across a 1200-km latitudinal gradient in Finland.

Regarding spatial dissimilarity, environmental filtering had a larger imprint than what random processes had. For butterflies and moths, environmental filtering shifted from being primarily associated with land cover to being primarily associated with climate indicating a likely regime shift along with warming climate. Regarding temporal dissimilarity of bird and butterfly communities, the imprints of environmental filtering and random processes varied between monitoring sites. A conventional stationary model was unable to track such site-specific processes. The imprints did not change linearly along a latitudinal gradient.

Our results demonstrate that accounting for non-stationarity in community dynamics is needed to pinpoint temporal shifts and spatial variability in the drivers of community change. Should we assume that community change is driven by the same primary forces at all times and everywhere, then we will fail to detect the real local and contemporary drivers of change, and risk applying the wrong corrective measures.