The evidence and global extent of alternative stable states in forest leaf phenology

Whether forests are composed of evergreen or deciduous species largely affects biogeochemical cycles and the functioning, structure and biodiversity of ecosystems. These leaf phenology types may be self-promoted through positive plant-soil feedbacks, which may shift the distribution of forest types towards alternative stable states. However, we still lack empirical evidence of phenological alternative stable states and a spatial understanding of where they might be present. Here, we test the presence of alternative stable states using forest inventory data at the continental (North America and Europe) and global scale. We show that the distribution of forest leaf phenology types is bimodal, and demonstrate the presence of positive feedbacks in recruitment, growth and mortality. Data-driven simulations show that the observed positive feedbacks are sufficient and necessary to produce the alternative stable states, which also lead to hysteresis during ecosystem transition. Spatial random forest models further reveal hotspots of alternative stable states in evergreen-deciduous ecotones and the poleward range limit of forests, which appear largely driven by soil feedbacks. Given the close connection between forest leaf phenology and ecosystem biogeochemical processes, our insights on evergreen vs. deciduous alternative stable states inform our understanding of the distribution of forest biomes, allowing more accurate quantification of carbon turnover and terrestrial climate feedbacks.