Estimating crown base height for Scots pine by means of the 3D geometry of airborne laser scanning data

Crown base height (CBH) is an important factor in relation to several characteristics of the tree stock. This paper introduces approaches for estimating tree-level CBH from airborne laser scanning (ALS) data that make use of features of computational geometry. For that purpose, the concepts of Delaunay triangulations and alpha shapes were applied and compared with approaches based on analysing return frequencies and predicting CBH by linear regression. These approaches were evaluated using test data on a total of 185 Scots pine trees, of which 136 were of sawlog size, that were detected and delineated from ALS data with a density of approximately 4 returns m–2. The results suggest that variables based on the frequencies of crown returns within predefined height bins are the most accurate for estimating CBH. By combining the best CBH estimate with the estimated tree height in linear regression, a root mean squared error (RMSE) of 1.4 m (14%) was achieved when all study trees were considered. The estimation was generally less accurate for the trees smaller than those of sawlog size. Although the accuracy of estimating CBH is lower using the three-dimensional (3D) geometry approaches presented here, they are considered to have potential for further development.