Abstract
<title>Abstract</title> <p>Ecosystem restoration is increasingly employed as a key response to anthropogenic biodiversity loss. While restoration generally leads to improvements in biodiversity and ecosystem functioning in degraded ecosystems, outcomes often are variable. Long-term experimental monitoring data are therefore essential for developing a science-based generalizable understanding of the average effects of restoration, the magnitude of variation in outcomes, and the mechanisms underlying this variation. Here, we quantify changes in butterfly communities over a period of up to 19 years using rigorous before-after control-impact experiment across 18 forestry-drained boreal peatland sites in Finland. Joint species distribution modeling reveals that peatland restoration consistently benefits mire-specialist butterfly species. Nonetheless, community responses to restoration are heterogeneous, with 60% of the observed variation explained by the amount of pristine peatland in the surrounding landscape. Strikingly, short-term responses were strongly correlated with long-term outcomes, indicating that restoration effect can be predicted just a few years after implementation. This result points to a compelling and previously underrecognized opportunity for early evaluation of restoration effectiveness, while reaffirming the critical value of long-term ecological experiments.</p>