Biodiversity decline with increasing crop productivity in agricultural fields revealed by satellite remote sensing

Increasing land-use intensity is a main driver of biodiversity loss in farmland, but measuring proxies for land-use intensity across entire landscapes is challenging. Here, we develop a novel method for the assessment of the impact of land-use intensity on biodiversity in agricultural landscapes using remote sensing parameters derived from the Sentinel-2 satellites. We link crop phenology and productivity parameters derived from time-series of a two-band enhanced vegetation index with biodiversity indicators (insect pollinators and insect-pollinated vascular plants) in agricultural fields in southern Sweden, with contrasting land management (i.e. conventional and organic farming). Our results show that arable land-use intensity in cereal systems dominated by spring-sown cereals can be approximated using Sentinel-2 productivity parameters. This was shown by the significant positive correlations between the amplitude and maximum value of the enhanced vegetation index on one side and farmer reported yields on the other. We also found that conventional cereal fields had 17% higher maximum and 13% higher amplitude of their enhanced vegetation index than organic fields. Sentinel-2 derived parameters were more strongly correlated with the abundance and species richness of bumblebees and the richness of vascular plants than the abundance and species richness of butterflies. The relationships we found between biodiversity and crop production proxies are consistent with predictions that increasing agricultural land-use intensity decreases field biodiversity. The newly developed method based on crop phenology and productivity parameters derived from Sentinel-2 data serves as a proof of concept for the assessment of the impact of land-use intensity on biodiversity over cereal fields across larger areas. It enables the estimation of arable productivity in cereal systems, which can then be used by ecologists and develop tools for land managers as a proxy for land-use intensity. Coupled with spatially explicit databases on agricultural land-use, this method will enable crop-specific cereal productivity estimation across large geographical regions.