Natural regeneration is increasingly recognized as a potentially cost‐effective strategy to reach ambitious forest landscape restoration targets, but rates of recovery are notoriously variable. We asked how well initial habitat conditions after cessation of agriculture predict forest recovery after nearly a decade. We aimed to provide land managers with general rules of thumb to assess when it is necessary to invest resources in active restoration, such as tree planting, to accelerate forest recovery.

Location: Coto Brus County, Puntarenas, Costa Rica.

We compiled data on initial vegetation structure, soil nutrients, prior land‐use history and surrounding forest cover at 13 sites. After 8.5 years, we measured vegetation indicators commonly used to assess forest recovery, namely amount of canopy closure and number and diversity of woody recruits.

Two variables, grass cover and canopy closure, measured 1.5 years after site abandonment, explained 47–87% of five of the six response variables after 8.5 years; recovery was faster in sites with lower grass cover and higher canopy closure initially. Waiting an additional year to measure initial vegetation variables did not improve model fit. Time since the original forest was cleared explained 62% of change in canopy cover, whereas percentage of surrounding forest cover, length of pasture use and soil variables explained minimal additional variation.