Major disturbances test resilience at a long‐term boreal forest monitoring site

1. The impact of disturbances on boreal forest plant communities is not fully understood, particularly when different disturbances are combined, and regime shifts to alternate stable states are possible after disturbance. A long‐term monitored semi‐natural forest site subject to intense combined storm and bark beetle damage beginning in 2005 provided an opportunity to investigate the postdisturbance development of the vegetation community. Previous studies suggest that a shift from Picea abies to Fagus sylvatica domination was possible. 2. We analyzed pre‐ and postdisturbance vegetation data to investigate to what extent vascular plant species abundances, diversity, traits, and community composition have changed. We were particularly interested in differences between remaining apparently unaffected areas (potential refugia) and disturbed areas, and in signs of consistent change over time in community composition in response to disturbance that could indicate an impending regime shift. 3. We found that the vegetation community present in the refuge areas has remained substantially intact through the period of disturbance. Nonrefuge areas diverged from the refuges in community composition and showed increased taxonomic and functional diversity. Despite this, and an increase in deciduous tree species (particularly F. sylvatica), P. abies has shown strong postdisturbance regeneration. The refuges may be important in the apparent ongoing recovery of the disturbed areas to a P. abies‐dominated state similar to that found predisturbance. This fast recovery is interpreted as evidence of a system resilient to a potential shift to a deciduous‐dominated state. 4. Synthesis: Our results show that even powerful combined disturbances in a system with multiple stable states can be insufficient to initiate a regime shift. Resilience of the P. abies‐dominated forest community is increased by the survival of refuge areas functioning as a form of ecological memory of the previous ecosystem state. The results also demonstrate the value of data generated by long‐term monitoring programs.