Fig. 5.

Uncertainty of AGB map, taking the error of the random forest mapping and the error of the lidar model. Example 1 shows how some lidar scenes show in the uncertainty map, because their prediction varies a lot depending on whether or not they were part of the training data for different iterations (a). Scatter plots of test samples based on the leave-one-scene-out cross validation: height (b), AGB form H (c)