Disentangling community assemblages to depict an indicator of biological connectivity: A regional study of fragmented semi-natural grasslands

Under the current context of global change that largely threatens overall biodiversity in increasingly fragmented landscapes, more insights are needed into the drivers of biological connectivity between communities (i.e. the flow of species among a set of local communities responding to landscape structure). This study aims at estimating an indicator of regional biological connectivity of semi-natural grasslands from the expected correlation between the degree of community assemblage (i.e. composition similarity) and landscape features directly related to dispersal among local communities. Large-scale plant distributions characterizing semi-natural grassland communities were gathered from the atlas of Brittany flora (NW France; UTM grid of 10 km x 10 km). The analysed variables were computed considering the focal UTM square and its immediate neighbours, and the modelling encompassed different regression techniques accounting for spatial autocorrelation [Simultaneous Autoregressive (SAR) error models] and non-stationarity [Geographically Weighted Regressions (GWR)]. The degree of community assemblage allocated to biological connectivity was 10.4% (adjusted-R2); it was mainly correlated to decreasing hedgerow length and secondly to increasing structural connectivity of semi-natural grasslands once spatial autocorrelation was accounted for. The estimation of the indicator of biological connectivity (0.05 ± 0.01 from the SAR models) were improved when considering non-stationarity issues, particularly for the Eastern part of Brittany (up to 0.12 in terms of biological connectivity). Overall, the proposed indicator and estimation methodology represent a step ahead in connectivity analysis at the community level, potentially relevant in the detection of hotspots of biological connectivity which can help buffer current large-scale biodiversity threats due to global change.