Estimating average tree crown size using high-resolution airborne data

Abstract

Tree crown size is a key parameter of tree structure that has a variety of uses, including assessment of stand density, tree growth, and amount of timber volume assessment. Remote sensing techniques provide a potentially low-cost alternative to field-based assessments, but require the development of algorithms to easily and accurately extract the required information. This study presents a method for average crown diameter estimation on a plot level based on high-resolution airborne data. The method consists of the combination of a window binarization procedure and a granulometric algorithm. This approach avoids the complicated crown delineation procedure that is currently used to estimate crown size. The method was applied to a spruce mountain forest and was verified on 23 reference plots. The method achieved best results of R 2 =76% [ RMSE=0.37 m (11.2% of the observed mean)] and R 2 =79% [ RMSE=0.49 m (16.7% of the observed mean)]. The study investigates the dependence of the algorithm results on the sun altitude of each image, and determines the optimal combination of spectral bands from hyperspectral airborne images for the application of the method.