Reconstructing 450 Years of Pollarding Events in Spanish Deciduous Oak Woodlands Using Machine Learning

Pollarding, the practice of pruning tree branches at a specific height, has been crucial for managing open forests in Europe. This practice has supported the persistence of highly biodiverse open woodlands featuring ancient trees. Understanding historical management patterns is essential for interpreting past socioeconomic conditions and developing strategies to mimic traditional practices for biodiversity conservation. Current methods for reconstructing past management in pollarded forests often rely on techniques for large-scale forest disturbances, which may be suboptimal for detecting short-term perturbations like pollarding. To address this gap, we applied a random forest algorithm to detect pollarding events using tree-ring traits, reconstructing the multi-centennial management history of four deciduous oak dehesas in northern Spain. Our analysis revealed that short-term changes in latewood were the most reliable indicator of pollarding events. Pollarding typically reduced latewood production for about three years, with the most pronounced declines occurring toward the end of the pollarding effect period. Pollarding patterns underwent a major shift starting in the last third of the 20th century. Key historical decades of both high and low pollarding pressure were consistently observed across the studied dehesas. These findings enhance our understanding of these unique ecosystems and offer critical insights for their conservation.