Geometrically explicit description of forest canopy based on 3D triangulations of airborne laser scanning data

Geometrically explicit parameterizations of the locations, orientations and properties of trees and canopy elements are useful for forest ecosystem modeling. Detailed reconstructions based on real tree geometry are, however, hard to obtain based on conventional field measurements and sampling. We describe and evaluate an alternative approach to reconstruct the forest canopy from sparse, leaf-off airborne laser scanning (ALS) data with a wall-to-wall coverage. The approach employs computational geometry and topological connectivity to generate filtrations, i.e., ordered sets of simplices belonging to the three-dimensional (3D) triangulations of the point data, and numerical optimization to select the set of simplices with a quasi-optimal relationship with field measured forest biophysical attributes. The approach was evaluated by predicting the quantities and spatial patterns of biomass-related forest attributes according to the characteristics of the filtration. When the filtration parameters were optimized for 245 sample plots of 300 m(2) located in southern boreal forest in Finland, the coefficients of determination (R-2) between total volumes of the filtrations and basal area, stem volume, total above-ground biomass, and canopy biomass were 0.93, 0.87, 0.87, and 0.62, respectively. Considerably less accurate results (R-2 = 0.44-0.64) were obtained when the filtration parameters were predicted with a limited number of the calibration field plots. However, these accuracies could be obtained with modest field training data of 20-30 plots. The proposed approach is a compromise between the parameterization of the forest scene by artificial tree/crown level turbid media and realistic 3D models. The results particularly suggest that obtaining coarse wall-to-wall descriptions does not require separate data acquisitions, but may be based on data existing due to previous practical inventories. (C) 2015 Elsevier Inc All rights reserved.