Global Natural and Planted Forests Mapping at Fine Spatial Resolution of 30 m

The expansion of planted forests often encroaches upon natural forests, leading to numerous environmental and social problems. Mapping natural and planted forests is crucial for the monitoring, management, and conservation of these invaluable forest resources. However, global mapping of natural and planted forests at fine spatial resolution remains an unaddressed need. Here, we generated more than 70 million training samples from dense Landsat images and fed them to a random forest classifier (RF). Our dataset achieved an impressive overall accuracy of 85% when validated against reference data.

人工林（planted forests）的扩张时常侵占天然林（natural forests），进而引发诸多环境与社会问题。对天然林与人工林进行精准制图，对于这类珍贵森林资源的监测、管理与保护至关重要。然而，获取精细空间分辨率的全球天然林与人工林制图数据，仍是一项尚未得到满足的需求。本研究依托高密度陆地卫星（Landsat）影像生成了超7000万条训练样本，并将其输入至随机森林分类器（random forest classifier, RF）中。经参考数据验证后，本数据集的总体精度达到了令人瞩目的85%。