Diverging restoration pathways for overstory and understory communities in a Mediterranean‐climate riparian ecosystem

The classic restoration ecology model of ecosystem recovery predicts that restoring the initial conditions of a formerly degraded site will facilitate recovery and convergence with a reference site. Few restoration studies have long-term longitudinal data to evaluate recovery trajectories, which typically vary among different aspects of ecosystem structure and composition. We used repeat surveys to evaluate if the trajectory of vegetation structure and community composition in restored and reference forests converged from recovery of restored forest or degradation of reference forest, and whether patterns differed for forest overstories and understories. We measured vegetation communities 9-16 years after prior surveys and 21-30 years after planting in 11 restored and 8 reference forest sites along a 100-km span of the Sacramento River, California. We used generalized linear mixed models and nonmetric multidimensional scaling to evaluate convergence in overstory structure, overstory composition, understory cover, and understory composition. Restored forest structure became more similar to reference forests over time, converging on a basal area of ca. 20 m² ha^-1 around 25 years and densities of stems around 35 years. In contrast, overstory species composition remained distinctive, primarily due to the different initial tree-species composition and limited overlap in new species recruitment in restored and reference forests. Restored understories had lower native and higher exotic cover than reference sites in both surveys. Restored understory composition became more similar to reference sites as shade-intolerant exotic species cover declined. However, understory composition in restored sites remained distinct from reference sites due to the increased woody cover, primarily from the invasive, shade-tolerant shrub Rubus armeniacus, which increased from 2% to 23% mean site cover between the two surveys. Path analysis showed that in restored forests, R. armeniacus cover was negatively related (R = -0.51) to native understory cover, suggesting that it impedes forest recovery. Our results suggest that this system has converged with reference forest in overstory structure but not yet in overstory or understory composition. We found substantial degradation of restored forest understories over time due to R. armeniacus spread; redirecting this trend would require substantial, ongoing intervention.