Alternative vegetation states and limited reversibility: Insights from a novel grazing experiment

Grazing by large herbivores can drive grasslands toward contrasting vegetation states. However, the reversibility of these changes remains uncertain. The reintroduction of grazing into long-term exclosures, and how these trajectories compare with newly established exclosures, are still poorly understood. We combined a 29-year grazing exclusion with adjacent continuously grazed savannas in the Río de la Plata grasslands to test whether vegetation changes induced by grazing exclusion and reintroduction are reversible at medium time scales. In a five-year field experiment, we monitored vegetation trajectories under four treatments: long-term exclusion, continuous grazing, exclusion followed by grazing, and grazing followed by exclusion. Results revealed that long-term exclusion and continuous grazing persisted as distinct states, supporting the existence of alternative stable states. New exclosures converged toward the vegetation structure of long-term ungrazed sites, after five years of exclusion that promoted accelerated shrub encroachment. In contrast, grazing reintroduction triggered non-linear changes, and did not recover the typical structure of grazed grasslands, but instead settled into an intermediate state characterized by persistent bare soil and the absence of prostrate grasses. The lack of recovery was constrained by the low abundance of prostrate species in the soil seed bank and their limited colonization capacity. These asymmetrical trajectories suggest ecological thresholds and hysteresis that constrain system reversibility. Our findings provide empirical support for the nonlinear dynamics of grasslands and highlight the value of the State-and-Transition Model guiding management and preventing transitions to degraded states.