Agricultural drivers of field margin plant communities are scale dependent

In recent decades, agricultural intensification has led to a strong decline in biodiversity. Field margins act as shelters and dispersal corridors for biodiversity in highly disturbed landscapes, and are critical to the maintenance of ecosystem services. However, they are also impacted by agricultural practices in neighbouring fields. Agricultural impacts are often studied at field to landscape scales, and rarely across biogeographic regions. One of the challenges in large-scale studies is the lack of standardized monitoring schemes including both biodiversity and accurate estimation of agricultural practices. Here, we take advantage of a national monitoring scheme in 462 sites in France, to assess the effects of agricultural practices on field margin flora at different extents and resolutions. We used spatial simultaneous autoregressive and generalized dissimilarity models to assess the response of plant richness and composition to climatic, soil and landscape conditions, and to agricultural (fertilization, herbicides) and margin management drivers. Analyses were repeated at the site-level, 40 and 75 km resolutions, and at regional and national extents. We found that the impact of agricultural practices on species richness was most important at the site-level, whereas climate and crop diversity became more important at the 75 km resolution. Compositional variations responded differently, with climate being more important at the site-level, and fertilization and crop diversity at the coarsest resolution. There was a strong variation in the variance explained by models among regions, but climate effects were weaker within biogeographic units compared to the national level, and different agricultural practices stood out as influential in different regions, suggesting that the regional context is fundamental in determining plant community structure. To efficiently conserve biodiversity, we therefore recommend the implementation of agricultural measures adapted to each region.