Study into Reflection Spectra and Classification Possibilities of Forest Vegetation Species for the Autumn Period in European Part of the Russian Federation in the Spectral Range (0.4–1.0) μm

Laboratory studies of the leaves and needles reflection spectra from tree species characteristic of the European part of the Russian Federation (birch, oak, maple, linden, aspen, spruce, pine) were conducted in the autumn season within the spectral range of (0.4–1.0) μm with a spectral resolution of 2 nm. For each tree species, samples were taken from different trees and from various branches of a single tree. If yellow patches appeared on the leaves or if the foliage began to yellow, the reflection spectra changed significantly: the range of variability in the spectral characteristics for different samples of the same species expanded, and the step in the spectral dependence of radiance coefficient decreased in the spectral region of transition from visible to near-IR range. The reflection spectra measured between 6 September and 3 October were used to assess the probability of the tree species correct and incorrect classification. It was shown that, despite the increased variability in the spectral characteristics of leaf samples from certain species, the use of a neural network allows for good classification results (although slightly worse than in summer). The use of measurements in the spectral range of (0.4–1.0) μm with a resolution of 2 nm and a relative root mean square noise value of 1 % allows for potentially achieving a probability of correct tree species classification above 74 % with an incorrect classification probability below 5.4 % during the autumn period.