Global Forest Management Type Map at 10m Resolution for 2020 Part 2

Le résumé fourni par la source

This is an experimantal dataset that represents the 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 was collected and expanded with two additional classes: rubber plantations and tree crops. The VITO team trained a hybrid classification framework combining a deep learning and pixel-based classification approaches using Sentinel-1 and Sentinel-2 imagery for the year 2020. The resulting product provides a global wall-to-wall map of forest management types at 10 m spatial resolution. The map is spit into two parts and uploaded as 2 seperate records due to the zenodo size limitations. Part 2 includes Europe, Asia, Africa, Australia and Oceania. 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 (monoculture plantations); 53 - Agroforestry; 100- Other trees(e.g. trees in urban areas). Detailed class definitions are provided in a separate file. The modelling framework follows the methodology developed for global land cover mapping and described in the Copernicus Global Land Cover documentation. The work was funded by WRI.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.