Assessment of the classification accuracy of the Globeland30 Forest class for the temperate and tropical forests of Mexico

This paper presents an assessment of the classification accuracy of the 2000 and 2010 GlobeLand30 (GL30) Forest land cover class for the forests in Mexico at the national level. The rich diversity of temperate and tropical forests in Mexico provides an ideal stage for this assessment. As validation data set, we used thousands of field sampling sites with detailed forest cover information from two sampling cycles of the national forests inventory of Mexico. We defined 9 distinct types of forests and assessed the GL30 Forest class classification accuracy for each of them. This accuracy was higher in tropical forests than in temperate forests (around 90% vs. around 77%). The lowest accuracy values (around 73%) were in the tropical and temperate forest with dominance of deciduous tree species. The largest number of wrongly classified sampling sites covered by forests fell in three GL30 classes: Shrublands, Grasslands, and Cultivated. In the temperate forests, almost 100% of the wrongly classified sites fell into the first two land covers. The more specific tropical forest types defined in our study showed distinct patterns of distribution of the misclassified sites between these three GL30 land cover classes.