The use of hyperspectral remote sensing for mapping the age composition of forest stands

Abstract

The paper deals with the issue of mapping the age composition of stand groups using hyperspectral imagery acquired by the AISA Eagle VNIR sensor in the Bílý Kríž locality in the Moravian-Silesian Beskids Mts. An object-oriented approach was employed through segmentation and subsequent classification by means of Nearest Neighbour (NN) algorithm in the environment of eCognition Developer 8 and artificial neural network (ANN) classification provided by ENVI 4.7 software. Because of the dominant occurrence of Norway spruce (Picea abies [L.] Karst.) monocultures in the studied locality the work focuses primarily on the distinguishability of two selected age classes of Norway spruce (10-20 years and 70-80 years). It studies possibilities of a more detailed age estimation of stand groups aged from 10 to 80 years based on the classification into the boundary classes, which shows similarity to dithering based on random algorithm. Comparison with the outline map of the Forest Management Plan shows a correlation (r2 = 0.83) between the spectral characteristics of Norway spruce stands and their age composition.