Global Forest Management Type Map at 100m resolution for 2020

This is an experimantal dataset that represents an updated version of the global forest management layer for 2015 developed by Lesiv et al. (2022). It is based on a new training dataset for the year 2020 and was collected and expanded with two additional classes: rubber plantations and tree crops. The VITO team trained a hybrid classification framework combining deep learning and pixel-based classification approaches using Sentinel-1 and Sentinel-2 imagery for the year 2020 at the resolution of 10m then aggregated to a 100m. The resulting product provides a global wall-to-wall map of forest management types at 100 m spatial resolution. Tree cover mask comes from the LCFM annual land cover map 2020. The map distinguishes the following classes: 0 – No tree cover 11 - Naturally regenerating forests without any signs of management, including primary forests; 20 – Naturally regenerated forests where there are clearly visible indications of human activities, such as selective logging, shifting cultivation, etc.; 31 – Planted forest 32 – Plantation Forest 33 – Rubber plantation 40 – Oil palm plantations 50 – Tree crops (mono plantations) 53 - Agroforestry 100- Other trees (e.g.) Detailed class definitions are provided in a separate file. The original 10m maps could be found here: The modeling framework is based on the methodology developed for global land cover mapping and described in the Copernicus Global Land Cover documentation. The work was funded by WRI.