Precipitation legacies reduce vegetation productivity and species diversity in semi-arid sandy grassland

Abstract Precipitation legacy effects (PLEs) profoundly alter the recovery trajectories of semi-arid grasslands under global climate change, necessitating mechanistic quantification for accurate climate risk assessment in these vulnerable ecosystems. Based on a seven-year precipitation simulation experiment followed by an in-situ natural recovery study in a semi-arid sandy grassland in Inner Mongolia, China, we characterized PLEs across multiple ecological hierarchies and varying precipitation patterns using data from the final treatment year and the first post-treatment year. Our results demonstrated that vegetation traits exhibited stronger PLEs than soil physicochemical properties. The magnitude of PLEs increased with higher functionalization (composition → productivity) and finer hierarchy (community → functional group), exceeding 50% when significant. Dry PLEs were generally stronger than wet PLEs; both exhibited bidirectional (positive/negative) performance, yet consistently showed an inverse relationship between vegetation traits and trait resilience. Mechanistically, PLEs of moderate wetting and extreme drying were primarily carried by vegetation-mediated information, whereas PLEs of moderate drying and spring drought legacies were mainly carried by soil-mediated material. Specifically, functional group composition served as the key information carrier: annuals primarily carried positive dry-PLEs and negative wet-PLEs, while perennials carried the opposite PLEs; and, the prevalent negative PLEs in community-level productivity and species diversity were specifically attributed to perennial grasses expansion after drying and annual forbs expansion after wetting. Soil available nutrients acted as the key material carrier, promoting PLEs in annuals via synergistic physicochemical pathways. Overall, both dry and wet PLEs generally impeded the vegetation recovery of the sandy grassland ecosystem, despite positive effects on certain finer-hierarchy ecosystem traits. We conclude that ignoring PLEs may lead to a severe underestimation of climate change risks in semi-arid ecosystems, particularly regarding their most sensitive components.