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.

人工林的扩张往往会侵占天然林的生存空间，进而引发诸多环境与社会问题。精准绘制天然林与人工林的分布图谱，对于这类珍贵森林资源的监测、管理与保护至关重要。然而，当前全球范围内实现精细空间分辨率的天然林与人工林制图仍存在未被满足的迫切需求。本研究基于密集的Landsat影像生成了超过7000万条训练样本，并将其输入至随机森林分类器（random forest classifier, RF）中进行训练。经参考数据验证，本数据集的总体分类精度可达85%，表现优异。

搜集汇总

背景与挑战

背景概述

该数据集提供了全球自然和人工林的30米空间分辨率映射，通过随机森林分类器处理超过7000万个Landsat图像训练样本，验证准确率达85%，旨在支持森林资源的监测、管理和保护。