Mapping tropical forests: implications and challenges for deforested landscapes and forest restoration. Examples from Zambia, Ecuador and Philippines : [paper for] XV World Forestry Congress, Coex, Soul, Republic of Korea, 2-6 May 2022

Within the framework established by international development agreements (e.g. Agenda 2030 or Paris Agreement), initiatives and programs for the protection, restoration and sustainable use of forests are being implemented worldwide (e.g. Forest Landscape Restoration, Bonn Challenge, REDD+). These programs require reliable and efficient forest cover monitoring tools, in order to measure their success or failure and to derive sound conclusions about their contribution to sustainable development. Despite the increasing quantity and quality of remote sensing-based global datasets and related National Forest Monitoring systems, establishing efficient monitoring systems in tropical regions is still facing operational challenges (e.g. lack of reference data, limitations to differentiate forest disturbance levels). This is especially relevant, as tropical forests and their multiple ecosystem services are shrinking fast if compared to other biomes. Thus, forest conservation and restoration programs require monitoring tools, which are able to differentiate levels of disturbance and stages of forest recovery. We conducted an extensive ground verification campaign and collected data on land use and forest disturbance history (~19,000 ha) through twelve provinces of three tropical countries (Zambia, Ecuador, Philippines). With this validation dataset, we analyzed time-series information (after the year 2000) of national forest maps, which are commonly used in international reporting (e.g. to FAO or UNFCCC), and two relevant global forest datasets (Global Forest Change and Tropical Moist Forests). We analyze forest dynamics of different forest types (reference, disturbed, regrowth), emphasizing the importance of in situ verification and time-series information to establish efficient forest cover monitoring systems in the tropics. Our results show that regrowth forests are especially prone to misclassification, which complicates the monitoring of success of restoration measures.