Abstract:It has been shown that height measurements obtained by airborne laser scanning (ALS) with high point density (>7-8 m´2) can be used to detect small trees in the alpine tree line-an ecotone sensitive to climate change. Because the height measurements do not discriminate between trees and other convex structures with positive height values, this study aimed at assessing the contribution of ALS backscatter intensity to classification of trees and non-trees. The study took place in a boreal-alpine ecotone in southeastern Norway and was based on 500 precisely georeferenced small trees and non-tree objects for which ALS height and intensity were derived from four different ALS acquisitions, representing different sensors, pulse repetition frequencies (PRF), and flying altitudes. The sensors operated at 1064 nm. Based on logistic regression modeling, it was found that classification into three different tree species ((1) spruce; (2) pine; and (3) birch)) and two different non-tree object types (objects with: (1) vegetated surface; and (2) rock) was significantly better (p < 0.001-0.05) than a classification based on models with trees and non-trees as binary response. The cause of the improved classification is mainly diverse reflectivity properties of non-tree objects. No effect of sensor, PRF, and flying altitude was found (p > 0.05). Finally, it was revealed that in a direct comparison of the contribution of intensity backscatter to improve classification models of trees and non-trees beyond what could be obtained by using the ALS height information only, the contribution of intensity turned out to be far from significant (p > 0.05). In conclusion, ALS backscatter intensity seems to be of little help in classification of small trees and non-trees in the boreal-alpine ecotone even when a more detailed discrimination on different species and different non-tree structures is applied.