In this master’s thesis, we assessed the applicability of unmanned aerial vehicle laser scanning (UAV-LS) for individual tree detection, estimation of tree-level attributes, and derivation of stand-level attributes in pure two-layered European beech stands in the Gorjanci Mountains. A total of 458 trees were measured in the field across 29 sample plots, and the results were compared with attributes derived from UAV-LS data. Individual tree detection was more successful in the upper canopy layer, whereas understory trees were often undetected or incorrectly detected. Tree height estimates differed from field measurements by an RMSE of 1.68 m (rRMSE = 5.1%), while estimates of diameter at breast height and tree volume showed larger deviations (DBH: RMSE = 10.59 cm; rRMSE = 23.7%; volume: RMSE = 1.66 m3; rRMSE = 57.5%). At the sample plot level, the error in estimating stand basal area was RMSE = 5.15 m2/ha (rRMSE = 18.5%), while the error in estimating growing stock was RMSE = 117.5 m3/ha (rRMSE = 24.6%). The number of trees was systematically underestimated from UAV-LS data, primarily due to the non-detection of understory trees. The results indicate that UAV-LS data are applicable to the analysis of the structure of European beech stands, particularly for estimating tree height, stand basal area, and growing stock. However, reference field measurements are still required for more reliable estimation of diameter at breast height, individual tree volume, tree number, and the understory layer. This study contributes to the development of modern approaches to digital forest inventory and confirms the applicability of UAV-LS data as a complement to conventional field measurements for monitoring forest condition and development.
|
