Intensive arable farming calls for resilient wild bee restoration amid severe abundance fluctuations.

Intensive arable farming threatens wild bees by reducing floral and nesting resources, increasing agrochemical exposure, and amplifying climatic stress. Ecological restoration using native wildflower mixtures is a promising intervention to recover pollinator diversity and the ecosystem service of pollination. However, restoration outcomes remain context dependent, and it is still poorly understood how restoration can be sustainably integrated into agricultural landscapes over time. We used a replicated, landscape-scale field experiment in East-Central Europe (Hungary) to test how spatial configuration, distance from restored plots, and landscape context jointly shape wild bee communities in intensively managed farmland. Twenty-four landscape plots were established across homogeneous and heterogeneous landscapes and assigned to two spatial configurations, compact wildflower fields and dispersed wildflower strips, with unplanted controls. We sampled the landscape plots within a 500-m radius using pan traps arranged in a grid over 4 years (2020-2023). Wild bee communities exhibited strong interannual fluctuations in abundance, species richness, and composition, with year emerging as the dominant driver and explaining more than one-third of variation in community composition. Bee abundance peaked in 2021-2022 but abruptly dropped in 2023, which may suggest heightened sensitivity of pollinator communities to the severe drought that occurred in 2022. Despite this strong temporal variability, wildflower fields in homogeneous landscapes supported significantly higher wild bee abundance than either homogeneous controls or fields in heterogeneous landscapes, suggesting that restoration benefits may be greater where floral deficits are most severe. Together, these findings indicate that pollinator restoration success depends on both spatial context and temporal variability. We suggest that restoration planning should prioritize long-term viability and resilience through sustained monitoring, stakeholder codesign, the use of diverse and locally adapted plant communities, and restoration schemes adapted to climatic variability. Restoration in homogeneous agricultural landscapes may provide benefits for pollinator communities and pollination services, while heterogeneous landscapes and semi-natural habitats remain essential for conserving less-degraded communities and rare species. Our results show that interannual variability can overshadow spatial configuration and landscape context, underscoring that even well-designed restorations must accommodate climatic and local variability to remain effective over time.