Accuracy assessment and uncertainty of 10-m resolution land use land cover maps at the local scale. Case: Talassemtane National Park, Morocco

Global land use land cover (LULC) maps are essential for monitoring forest dynamics and are a useful tool for any spatial planner. An increasing number of institutes are interested in producing such maps, especially with freely available high-resolution imagery and the emergence of new classification techniques. However, these maps should be submitted to an accuracy assessment for more accurate and effective use at the regional, national, and local scales. In this study, two products, the European Space Agency World Cover 2020 (ESA WC2020) and Sentinel-2 10 m Land Use/Land Cover Time Series (ESRI LC2020), were subjected to an accuracy assessment at the local scale, particularly in Talassemtane National Park (TNP). The results showed an overall accuracy of 56.90 % for the ESA product and 83.56% for the ESRI product. Forest cover is the most accurate land cover class for both products. In addition, area-based accuracy was used to adjust each class’s area and evaluate uncertainty. Finally, spatial agreement between the two products was achieved using change detection analysis, which showed a result of 70% at the local level. This work will be used in studies in the TNP area to distinguish between forest and non-forest land, particularly to monitor deforestation.