FRA collects data on forests every five years through country reports prepared by national correspondents who are nominated by government agencies responsible for forestry. When countries or territories do not nominate a national correspondent, estimated values for forest statistics are prepared by experts in FAO and published as desk studies.

In the FRA 2020 cycle, some desk studies were prepared through remote sensing analyses, using geospatial modules included in the FRA reporting platform and, if needed, through more analyses performed in the free cloud computing platform Google Earth Engine.

With one FRA geospatial tool, first an analysis is done of all the freely available tree cover and land use maps of the country at global and regional level. Next, a hybrid map is produced from a wall-to-wall comparison of all the global and regional maps available. A level of agreement of 75 percent is used, the percentage reflecting the probability of finding forest in the same pixel of every map considered.

If neither the hybrid map nor any individual global and regional map is deemed precise enough at country level, experts in the FRA team can prepare a national land cover map in the cloud computing platform Google Earth Engine. The national land cover map is obtained with a supervised classification of cloud-free mosaics through machine learning algorithms such as Random Forest or CART.

For some countries and territories without a national correspondent, the land use and land use change areas were assessed with Collect Earth using the same methodology as the FRA Remote Sensing Survey 2020.