Predicting the Plot Volume by Tree Species Using Airborne Laser Scanning and Aerial Photographs
Several studies have indicated that forest characteristics can be accurately predicted using airborne laser scanner (ALS) data, but there are very few studies in which species-specific forest characteristics have been estimated. This article compares two approaches for determining species-specific volumes at plot level by combining ALS data with aerial photographs. The first approach consists of two stages: (1) prediction of total volume using ALS data, and (2) assignment of this total volume to tree species by fuzzy classification and aerial photographs, in which three fuzzy classification methods were tested. In the second approach, volumes by tree species and the total volume are predicted simultaneously using a nonparametric k-most similar neighbor (k-MSN) method based on both ALS data and aerial photographs in one phase. The test area, located in Finland, consists of 463 sample plots. Species-specific volumes were estimated for pine, spruce, and the deciduous trees as a species group, total volume being the sum of the species-specific volumes. The k-MSN method produced considerably more accurate estimates for the species-specific volumes than any fuzzy classification method, the relative RMSEs for the volumes of pine, spruce, and deciduous trees being 45.50%, 61.98%, and 92.30%, respectively, and that for the total volume 23.86%.