Beschreibung:
<jats:title>Significance</jats:title>
<jats:p>Phytochemical diversity affects plant fitness and is the source of numerous medicines. Despite this, we know remarkably little about how phytochemical diversity is distributed across the plant kingdom and the environment. To address this challenge, we coupled untargeted metabolomics on 416 grassland vascular plant species across Switzerland with phylogenetic information, species distribution modelling, and ensemble machine learning to construct a framework for comprehensively predicting landscape-scale phytochemical diversity of both known and currently unclassified molecules. We show that phytochemicals diversity and identity can be predicted in the landscape as a function of phylogenetic information, as well as of climatic, topographic, and edaphic factors. We demonstrate that it is thus possible to map hotspots of phytochemical diversity and phytochemical endemism across the landscape.</jats:p>