Changes in the Fine Composition of Global Forests from 2001 to 2020

Abstract

Knowledge of forest management types is key to sustainable forest restoration practices, forest biomass assessment, and carbon accounting. However, there are no available global forest-management maps because of the spectral similarity of different forest management types. As such, we applied random forest and change detection algorithms to generate annual maps of 6 forest management types at a spatial resolution of 250 m from 2001 to 2020 including naturally regenerated forest (unmanaged and managed), planted forest (rotation of >15 years and ≤15 years), oil palm plantation, and agroforestry. In general, validation results on a point scale show that the overall accuracy is 86.82% ± 9.14%, indicating that our annual maps accurately represent global spatiotemporal variations in forest management types. Furthermore, we estimated the annual biomass carbon stock of different forest management types. The net expanded areas of planted forest, oil palm plantation, and agroforestry offset 59.56% of the loss of forest area and 77.13% of the loss of biomass carbon stock due to the decrease in the naturally regenerated forest. The decrease of managed natural regeneration forests, the expansion of planted forests with a rotation period of more than 15 years, and agroforestry resulted from reforestation practices, while the expansion of planted forests with a rotation period of less than 15 years and oil palm plantations resulted from the removal of part of agroforestry. Moreover, the expansion of planted forests with a rotation of less than 15 years (72.73%) dominates the global expansion of planted forests, and China has contributed 42.20% of this expansion. Our results are beneficial for nature solution-based climate change mitigation.