The pace and drivers of community change vary over space and time

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.

Methods
Birds

The bird monitoring data originates from 1149 transects surveyed in 1978-2020. Transects are 3-6 km long and visited once per survey year. Surveys are typically scheduled for June, with some variation in exact dates due to the latitudinally varying phenology of the breeding season. Not every transect is surveyed each year, whereas survey intensity (walking speed, survey methodology and width of surveyed area around a transect) is constant over surveys. The data has been curated and processed by the Finnish Museum of Natural History. In total 145 different bird species were detected during the monitoring. Surveyors counted abundance per species.

Butterflies

The butterfly monitoring data originate from 101 sites in agricultural landscapes surveyed in 1999-2020. Surveys are conducted by volunteers and the monitoring program is directed by the Finnish Environment Institute. Not every transect is surveyed each year. Surveys are conducted at least seven times per site in a survey lasting from May to August. Given latitudinal gradients in the length of the season, the survey period varies from 10 weeks in Northern Finland to 16 weeks in Southern Finland. The survey intensity of each single transect visit is held constant between sites and years and the species identification skill of the volunteers is high. In total 92 different butterfly species were detected during the monitoring. Surveyors counted abundance per species.

Moths

Moths were monitored by light traps at 99 sites in 1993-2018. The monitoring program is conducted under the National Moth Monitoring scheme (Nocturna) and directed by the Finnish Environment Institute. Traps are kept open from April to October depending on the moth activity period and emptied weekly by volunteers. The surveys in which moth traps had been kept open under six or over eight days were discarded to control for the varying sampling effort. Due to varying resource allocation and different moth activity periods, sites may have been visited at different times between years and sites. (Note that this sampling-related variation is explicitly accounted for in the model.) The data have been curated by monitoring coordinators. In total 908 different moth species were detected during the monitoring. Surveyors counted abundance per species.